nuclear regulatory commission contract nrc-02-97-009 · glass melting process that potentially can...
TRANSCRIPT
CNWRA 2000-05
Prepared for
Nuclear Regulatory CommissionContract NRC-02-97-009
Prepared by
Center for Nuclear Waste Regulatory AnalysesSan Antonio, Texas
September 2000
CNWRA 2000-05
GLASS MELT CHEMISTRY AND PRODUCT QUALIFICATION
Prepared for
Nuclear Regulatory CommissionContract No. NRC-02-97-009
Prepared by
Vijay JainYi-Ming Pan
Center for Nuclear Waste Regulatory AnalysesSan Antonio, Texas
September 2000
PREVIOUS REPORTS IN SERIES
WT-mhar Name- Date IssuedI 14 LUJLJLU-It . I ---
NUREG/CR-6666
NUREG/CR-575 1
CNWRA 97-0001
Survey of Waste Solidification Process Technologies
Hanford Tank Waste Remediation System High-LevelWaste Chemistry Manual
Hanford Tank Waste Remediation System FamiliarizationReport
In press
May 1999
July 1997
.Mi
ABSTRACT
Vitrified waste forms produced by melting high-level waste (HLW) with glass forming oxides, are being used
in the United States, France, England, Germany, Belgium, Japan, the Former Union of Soviet Socialists
Republic, and India as the preferred technology for the disposal of HLW. Both the vitrification technology
and the guidelines for disposal at aproposed geological repository impose stringent requirements for producing
an acceptable product. The glass composition designed for vitrification should meet (i) processing constraints,
such as viscosity, melting and liquidus temperatures, electrical conductivity, redox condition into the melt,
solubility ofnoble metals, and sulfur concentration; and (ii) product constraints, such as durability, crystallinity,
and glass transition temperature. This report provides a review of glass-melt properties and characteristics,
the basis and importance of the imposed constraints on the design of glass composition, and aspects of the
glass melting process that potentially can compromise worker safety. The review was conducted to assist the
U.S. Nuclear Regulatory Commission inthe development ofthe technical and regulatorytools for conducting
radiological safety assessments of the privatized vitrification facilities.
The review of the history of waste form development, and the information (data and models) on glass-melt
properties for various types of commercial and waste-glass compositions indicatethatthe behavior of glasses
is complex and cannot be expressed in simple terms for all wastes. The glass composition needs to be
designed uniquely for each waste type. The lessons learned, or the existing database of HLW glasses, can
be used as a guideline to narrow the scope of R&D needed to meet the requirements.
v
CONTENTS
Section Page
FIGURES.................................................................... xi
TABLES. ...................................................................... xvACRONYMS/ABBREVIATIONS ....................... xviiACKNOWLEDGMENTS .................. x,,,,. . xIX
EXECUTIVE SUMMARY ................. xxi
1 INTRODUCTION .............................. ,. 1-11.1 BACKGROUND .1-11.2 SCOPE OF REPORT . 1-1
2 HISTORY OF WASTE FORM RESEARCH, EVALUATION,AND SELECTION .. 2-12.1 HIGH-LEVEL WASTE TECHNOLOGY PROGRAM .2-12.2 PLUTONIUM IMMOBILIZATION PROGRAM .2-32.3 SELECTION OF A WASTE FORM FOR HANFORD HIGH-LEVEL
WASTE ........................................................ 2-52.4 CURRENT STATUS OF VARIOUS WASTE FORMS .2-62.5 REFERENCES ................ , 2-6
3 WASTE ACCEPTANCE CRITERIA FOR GEOLOGIC DISPOSAL . .3-13.1 PRINCIPAL WASTE ACCEPTANCE CRITERIA .3-13.2 REFERENCES ............... 3-4
4 GLASS FORMATION . . . 4-14.1 GLASS FORMATION PRINCIPLES .. 4-2
4.1.1 Zachariasen's Random Network Theory .4-24.1.2 Sun's Single-Bond Strength Criterion .4-3
4.2 GLASS FORMING SYSTEMS .. 4-34.2.1 Vitreous Silica Glass .4-34.2.2 Alkali Silicate Glass .4-34.2.3 Alkali-Alkaline Earth-Silicate Glasses .4-124.2.4 Alkali Borate Glasses .4-124.2.5 Alkali Borosilicate Glasses .4-144.2.6 Alkali Aluminosilicate Glasses .4-154.2.7 Alkali Aluminoborosilicate Glasses .4-154.2.8 Fe3+ in Alkali Silicate, Aluminosilicate, and Borosilicate Glasses .4-16
4.3 USE OF NONBRIDGING OXYGEN IN DEVELOPING NUCLEARWASTE GLASS PROPERTY MODELS . .4-17
4.4 SUMMARY ............................... 4-18
4.5 REFERENCES , ............................... 4-18
vii
CONTENTS (cont'd)
Section Page
5 ELECTRICAL CONDUCTIVITY ............................................ 5-15.1 ELECTRICAL CONDUCTIVITY LIMITS ............. .................. 5-15.2 ELECTRICAL CONDUCTIVITY IN GLASS MELTS ........ ............. 5-15.3 ELECTRICAL CONDUCTIVITY DATA AND MODELS ....... ........... 5-45.4 EFFECT OF MELTER CONDITIONS ON ELECTRICAL
CONDUCTIVITY ................................................ 5-85.4.1 Noble Metal Accumulation .............. ....................... 5-95.4.2 Formation of Conductive Species .......... ...................... 5-9
5.5 TANK WASTE REMEDIATION SYSTEM CONCERNS ....... ............ 5-95.6 SUMMARY .................................................. 5-95.7 REFERENCES ................................................ 5-10
6 GLASS MELT VISCOSITY ............................................... 6-16.1 MELT VISCOSITY LIMITS . ......................................... 6-16.2 VISCOSITY OF GLASS MELTS ............... ....................... 6-16.3 VISCOSITY DATA AND MODELS ............ ....................... 6-56.4 EFFECT OF MELTER CONDITIONS ON VISCOSITY ...... ............. 6-96.5 TANK WASTE REMEDIATION SYSTEM CONCERNS ...... ............. 6-96.6 SUMMARY .................................................. 6-96.7 REFERENCES ................................................. 6-10
7 GLASS TRANSITION TEMPERATURE ...................................... 7-17.1 GLASS TRANSITION TEMPERATURE LIMIT ........ .................. 7-17.2 GLASS TRANSITION TEMPERATURE MEASUREMENTS ..... ......... 7-17.3 GLASS TRANSITION TEMPERATURE DATA AND MODELS ..... ....... 7-47.4 EFFECT OF MELTER CONDITIONS ON GLASS TRANSITION
TEMPERATURE . .................................................. 7-67.5 TANK WASTE REMEDIATION SYSTEM CONCERNS ..... .............. 7-97.6 SUMMARY . ...................................................... 7-97.7 REFERENCES . .................................................. 7-10
8 LIQUIDUS TEMPERATURE ............................................... 8-18.1 LIQUIDUS TEMPERATURE LIMITS .................................. 8-18.2 LIQUIDUS TEMPERATURE MEASUREMENTS ........................ 8-18.3 LIQUIDUS TEMPERATURE DATA AND MODELS ...... ............... 8-28.4 EFFECT OF MELTER CONDITIONS ON LIQUIDUS TEMPERATURE ..... 8-128.5 TANK WASTE REMEDIATION SYSTEM CONCERNS ..... ............. 8-128.6 SUMMARY . .................................................... 8-128.7 REFERENCES . .................................................. 8-13
viii
CONTENTS (cont'd)
Section Page
9 REDOX .............. 9-19.1 REDOX LIMITS .. 9-1
9.2 THERMODYNAMICS OF REDOX REACTIONS . .9-19.2.1 Oxygen Ion Activity .9-49.2.2 Effect of Temperature on Redox Equilibrium ........ 9-69.2.3 Effect of Redox Couple Activity .9-79.2.4 Oxygen Activity in Glass Melts .9-8
9.3 REDOX STRATEGY AT HIGH-LEVEL RADIOACTIVEWASTE VITRIFICATION PLANTS .. 9-10
9.3.1 West Valley Demonstration Project Redox Control Strategy .9-119.3.2 Defense Waste Processing Facility Redox Control Strategy .9-13
9.4 REDOX ISSUES FOR TANK WASTE REMEDIATION SYSTEM .. 9-179.5 SUMMARY .. 9-19
9.6 REFERENCES .. 9-19
10 NOBLE METALS IN HIGH-LEVEL WASTE GLASS MELTS . .10-110.1 NOBLE METAL SOLUBILITY IN GLASS MELTS .10-110.2 NOBLE METAL BEHAVIOR IN THE MELTER .10-210.3 TANK WASTE REMEDIATION SYSTEM CONCERNS .10-410.4 SUMMARY .10-410.5 REFERENCES .10-5
11 SULFUR SOLUBILITY IN WASTE GLASSES . . .11-111.1 SULFUR SOLUBILITY LIMITS IN HIGH-LEVEL
WASTE BOROSILICATE GLASSES .. 11-1
11.2 SULFUR SOLUBILITY STUDIES .. 11-211.2.1 Effect of Redox .11-2
11.3 EFFECT OF MELTER CONDITIONS ON SULFUR SOLUBILITY.. 11-711.4 TANK WASTE REMEDIATION SYSTEM CONCERNS . .11-711.5 SUMMARY .. 11-811.6 REFERENCES .. 11-9
12 CHEMICAL DURABILITY .. 12-112.1 CATION RELEASE SPECIFICATION .12-1
12.1.1 Production Specification .12-112.1.2 Geologic Disposal Specification .12-1
12.2 GLASS CORROSION MECHANISM .12-212.3 GLASS DURABILITY MEASUREMENTS .12-5
12.3.1 Static Tests .12-612.3.2 Dynamic Tests .............................................. 12-9
ix
CONTENTS (cont'd)
Section Page
12.4 PARAMETERS AFFECTING LEACHING BEHAVIOR ...... ........... 12-1012.4.1 Effect of Surface Area to Volume Ratio .......................... 12-1012.4.2 Effect of Flow Rates ............ ............................. 12-1012.4.3 Effect of Temperature ........................................ 12-1312.4.4 Effect of Glass Composition ......... ......................... 12-1512.4.5 Effect of Contact Solution .......... .......................... 12-20
12.5 CHEMICAL DURABILITY MODELS .............. .................. 12-2412.5.1 General Chemical Durability Models ............................ 12-2412.5.2 Long-Term Chemical Durability Models ......................... 12-34
12.6 GLASS BEHAVIOR IN THE REPOSITORY ENVIRONMENT .... ........ 12-3612.7 TANK WASTE REMEDIATION SYSTEM CONCERNS ....... .......... 12-3812.8 SUMMARY .................................................... 12-3812.9 REFERENCES .................................................. 12-39
13 PHASE STABILITY ........... .. 13-113.1 PHASE STABILITY SPECIFICATION .. 13-113.2 PHASE SEPARATION .. 13-1
13.2.1 Immiscibility Boundary Measurements .13-313.2.2 Phase Separation in High-Level Waste Glasses .13-3
13.3 CRYSTALLIZATION .. 13-713.3.1 Crystallization Kinetics Model .13-813.3.2 Time-Temperature-Transformation Diagram Measurements .13-913.3.3 Time-Temperature-Transformation Diagrams for High-Level
Waste Glasses .13-913.4 EFFECT OF PHASE STABILITY ON GLASS DURABILITY . . 13-11
13.4.1 Effect of Phase Separation . 13-1113.4.2 Effect of Crystallization . 13-12
13.5 TANK WASTE REMEDIATION SYSTEM CONCERNS . . 13-1313.6 SUMMARY .. 13-1313.7 REFERENCES .. 13-14
14 SUMMARY ....... 14-1
x
FIGURES
Figure Page
4-1 Specific volume-temperature diagram for a glass-forming liquid ..................... 4-14-2 Schematics of the silica glass network .......... .............................. 4-12
4-3 Thermal expansion coefficient and density of Na2O-B2 03 glasses ..... .............. 4-134-4 The fraction N4 of 4-coordinated boron atoms in alkali borate glasses as a function
of mol% alkali . 4-14
4-5 Fraction of tetrahedrally coordinated borons (N4) as a function of mol Na2O/B2 03and Si02 ratios determined using nuclear magnetic resonance and moleculardynamics simulation . 4-15
4-6 The ratio of Fe2+/Fet., in (20-x) K20xAl203g80SiO2 containing approximately1 mol% Fe203 . 4-16
5-1 Conductivity (a) isotherms for 25R20-1 OA1203 mixed-alkali melts against K20/M20 at fivetemperatures ............................... 5-2
5-2 Schematic of an electrical resistivity apparatus ............................... 5-35-3 Predicted component effects on electrical conductivity at 1,150 0C relative to the
HW-39-4 composition, based on the first-order mixture model using mass fractions ..... 5-6
5-4 Conductivity as a function of alkali content at 1,150 CC ............................ 5-8
6-1 Variation of the viscosity of a common soda lime silica glass with temperature .......... 6-26-2 Viscosity of melts in binary silicate and germanate systems as a function of
alkali oxide (M 20) concentration . ............................................ 6-26-3 Isokom (equal viscosity) temperatures for xNa2O (20-x) K20-80 Si0 2 glasses .... ...... 6-36-4 Schematic of a rotating viscometer . ........................................... 6-46-5 Predicted component effects on viscosity at 1,150 °C compared to the HW-39-4
composition, based on the first-order mixture model using mass fractions ..... ........ 6-8
7-1 Determination of transition temperature using differential scanning calorimetry .... ..... 7-27-2 Determination of transition temperature using dilatometry .......................... 7-3
7-3 Predicted component effects on transition temperature relative to the HW-39-4composition, based on the first-order mixture model using mass fractions ..... ........ 7-7
7-4 Typical canister cooling profiles in the West Valley Demonstration Project ..... ........ 7-9
8-1 Liquidus temperatures of major crystalline phases in compositional variability studyglasses .8-3
8-2 Predicted versus measured liquidus temperature of clinopyroxene for the first-ordermixture model .8-4
8-3 Predicted versus measured liquidus temperature of spinel for the first-ordermixture model. 8-5
8-4 Predicted versus measured liquidus temperature of Zr-containing crystals for thefirst-order mixture model. 8-5
xi
FIGURES (cont'd)Figure Page
8-5 Predicted component effects on liquidus temperature of clinopyroxene relativeto the HW-39-4 composition, based on the first-order mixture modelusing mass fraction . ........................................................ 8-7
8-6 Predicted component effects on liquidus temperature of spinal relative to theHW-39-4 composition, based on the first-order mixture model using mass fraction ..... 8-8
8-7 Predicted component effects on liquidus temperature of Zr-containing crystals relativeto the HW-39-4 composition, based on the first-order mixture model usingmass fraction ....... 8-9
9-1 Redox chemistry in WV-205 melt at 1,150 0 C superimposed on theSRL-131 melt system ............. 9-4
9-2 The distribution of redox states of various multivalent elements inSRL-1 31 containing model nuclear waste at1,150 'C as a function of the imposed oxygen fugacity .............. .............. 9-5
9-3 Effect of composition of alkali (M) silicate melts on the concentration ratioFe3+/Fe2+ for melts equilibrated with air at atmospheric pressure and 1,400 'C;total Fe in the melt is less than 0.5 percent ........... ................. 9-6
9-4 The temperature dependence of the Fe3+-Fe2 ' equilibrium in melt composition(SRL-131) .............................................................. 9-7
9-5 The dependence of the redox state of iron on both the imposed oxygen fugacity andthe melt temperature in West Valley glass. Circles represent results obtained at1,250 'C, while squares represent those at 1,050 'C .............................. 9-8
9-6 Effect of excess total organic carbon on the Fe2̀ /Fe3 ' ratio at a fixed NO3 content ...... 9-129-7 Effect of H20 on the Fe2"/Fe 3 ' at a fixed total organic carbon/NO 3 ratio ..... ......... 9-129-8 The dependence of log Fe2"/Fe 3 1 on the index of feed oxidation: a comparison
between slurry fed ceramic melter and pilot-scale melters . 9-149-9 West Valley Demonstration Project redox forecasting model revised after
addition of nitrites . .............................................. 9-159-10 Comparison of scale glass melter and integrated Defense Waste Processing Facility melter
system campaigns with crucible redox data ............... ..................... 9-169-11 Comparison of crystalline phases with slurry (F-N) and glass redox ...... ........... 9-169-12 Effect of feed chemistry and copper content on the likelihood of formation
of insoluble copper precipitates during vitrification .......... .................... 9-18
11-1 Sulfurredox/solubilitysystematics inWVF-1 7993 (solid line), in SRL-131 (dotted line) andin SRL-i 31 + 10 wt% Fe (dashed line) as a function of the imposed oxygen fugacity in anatmosphere of 15 vol% SO2 at 1,150 °C . 11-2
11-2 S03 distribution in soluble gall, glass, and volatilized fraction as a function ofS03 content . 11-3
11-3 S03 distribution in soluble gall, glass, and volatilized fraction as a function ofA12 03 content ........................................................... 11-4
xii
FIGURES (cont'd)
Figure Page
11-4 SO3 distribution in soluble gall, glass, and volatilized fraction as a functionof P205 content .......................................................... 11-4
11-5 Sulfate retention in glass versus (CNBO)2/CBO, where CNBO and CBO arecalculated mol concentrations of nonbridging oxygen and bridging oxygen .... ........ 11-6
12-1 (a) Glass dissolution mechanism, (b) Schematic of surface layer on leached glass ....... 12-312-2 Effect of pH on the rate of extraction of silica from fused silica powder at 80 CC .... ... 12-412-3 Effect of pH on the extraction of soda and silica from a Na2O 3SiO 2 glass at 35 0 C ..... 12-412-4 Concentration versus (SAV)-time for release of B from (a) SRL-131 glass and
(b) SRL-202 glass at 90 0C at (0) 10, (U) 2,000, and (*) 20,000 ml . 12-1112-5 Leachate pH values (25 'C) versus reaction time for (a) SRL-1 31 glass and
(b) SRL-202 glass reacted at 90 0C at (0) 10, (U) 2,000, and (*) 20,000 m 1 . 12-12
12-6 Normalized release rates of sodium and silicon as a function of leachantflow rates . ...................................................... 12-14
12-7 Normalized total mass loss versus leach time for various flow rates ..... ........... 12-14
12-8 Normalized elemental mass loss versus temperature for different leaching durations ... 12-1512-9 Summary diagram indicating reported activation energies for individual reaction
processes and overall activation energies for high-level waste glass studies .. 12-1612-10 Leaching (MCC-3 test, 90 °C) of WV-205 glass as a function of glass redox state .... 12-18
12-11 Interpolated surface showing the dependence of MCC-3 boron concentrationon test time and amount of Si0 2 added to WV-205 ............................. 12-18
12-12 Interpolated surface showing the dependence of MCC-3 boron concentrationon test time and amount of ZrO2 added to WV-205 ............................ 12-19
12-13 Interpolated surface showing the dependence of MCC-3 boron concentrationon test time and amount of A1203 added to WV-205 ............................ 12-19
12-14 The Si leach rates (mg/m/d) of glass WV-205 in deionized water, 0.1 M KNO3,0.1 M Ba(NO3 )2 and 0.1 M CsNO3 solutions at neutral pH for up to 200 d of leaching 12-23
12-15 Cumulative normalized leach concentration for boron versus time for theWest Valley Demonstration Project (WVDP) glass in various solutions .12-24
12-16 Normalized leach rate for various elements versus leachate pH after the firstsolution replacement for West Valley Demonstration Project (WVDP) glass .12-25
12-17 (a) Linear regression plots of the AGhyd term versus silicon release to the leachantin a 28-d static leach test. The AG,.d term has been adjusted for pH changes, asdiscussed in the text. (b) Linear regression plots of the AGhd term versus boronrelease to the leachant in a 28-d static leach test. The AGhyd term has been adjustedfor pH changes, as discussed in the text. 12-29
12-18 Predicted component effects on MCC-1 release relative to the HW-39-4 composition,based on the first-order mixture model using mass fractions fitted to the reduceddata set .. 12-32
xiii
FIGURES (cont'd)
Figure Page
12-19 Predicted component effects on PCT B release relative to the HW-39-4 composition,based on the first-order mixture model using mass fractions ...... ................ 12-33
13-1 Liquid immiscibility in a binary eutectic diagram of components A and B. XB isthe mole fraction of B ..................................................... 13-2
13-2 Miscibility limits in a variety of binary silicate systems involving monovalent (X+)and divalent (M2+) cations .............................. 13-4
13-3 Prediction of phase separation using various normalized submixtures of(a) Na2O-B203-SiO2, (b) (Na2O+Li2O)-B203 -SiO2, and (c) Na2O(equivalent)-B 2 03 -SiO2. . 13-5
13-4 Compositional distinction between homogeneous and phase-separatedglasses with 95-percent confidence ellipsoids ............................... 13-6
13-5 Variation of nucleation rate and crystal growth rate with temperature .13-713-6 Time-temperature-transformation diagram for the Defense Waste Processing
Facility Blend glass .13-1013-7 Time-temperature-transformation diagrams for 2 and 4 vol% crystallization of
the oxidized West Valley Demonstration Project Reference 6 Glass. Horizontalbars represent 95-percent confidence intervals .13-11
13-8 Time-temperature-transformation diagrams for (a) isothermal and(b) nonisothermal crystallization for a simulated Hanford glass with variousvolume fraction of spinel .13-12
xiv
TABLES
Table Page
2-1 Candidate waste forms considered for geologic disposal of high-level waste ..... ....... 2-22-2 Initial screening criteria for Pu immobilization waste forms ......................... 2-32-3 Final ranking of waste forms for Pu immobilization ............................... 2-4
3-1 Waste acceptance preliminary specifications .................................... 3-2
4-1 Single-bond strengths for oxides ............... .............................. 4-44-2 Commercial glass compositions in oxides (wt%) ................................. 4-5
4-3 Compositions of selected candidate waste glasses (mol%) ........ .................. 4-6
5-1 Composition region (mass fraction) studied by Hrma et al. (1994) ...... ............. 5-45-2 Regression coefficients for In I1150 -. 5-55-3 Compositional ranges of six multicomponent systems and their viscosity and
conductivity ranges at 1,150 C .5-7
6-1 Regression coefficients for In rj1150- . -. .. ....---------- 6-7
7-1 Transition temperature for Defense Waste Processing Facility high-level wastecompositions using fiber elongation method .................................... 7-4
7-2 Defense Waste Processing Facility projected compositions ......................... 7-57-3 Regression coefficients for transition temperature ................................ 7-57-4 Effect of Fe2"/Fe 3 1 on transition temperature (C) ................................ 7-87-5 Effect of isothermal and nonisothermal heat treatment on transition temperature (C) .... 7-8
8-1 Regression coefficients for liquidus temperature ................................. 8-48-2 Multicomponent crystallinity constraints and their correlation with liquidus temperature . 8-108-3 Composition regions (in mass fractions of components) for Defense Waste
Processing Facility and SP and SG glasses ..................................... 8-108-4 Regression coefficients for liquidus temperature ................ ................ 8-11
9-1 Redox couples in nuclear waste glasses ........................................ 9-29-2 Effect of various redox compounds in closed crucible tests ......... ............... 9-10
10-1 List of noble metal isotopes present in high-level waste . 10-110-2 Noble metal concentration in various high-level waste glasses ...................... 10-210-3 Amount of noble metals processed through integrated Defense Waste
Processing Facility melter system ............ ................................ 10-3
11-1 SO4 concentration limits for A, B, and C low-activity waste feed envelopes .... ....... 11-111-2 Formation of salt layer as a function of redox in SO-1-10 glass ..................... 11-711-3 Formation of salt layer as a function of redox in glass containing 9 wt% CaO .... ...... 11-8
xv
TABLES (cont'd)
12-1 Summary of leach test methods ............... .............................. 12-712-2 Leaching experimental results after varying composition ......................... 12-2112-3 Primary basis set of partial molar hydration free energies, AGJ, for glass in
oxidized pH regimes >7 . .................................................. 12-2612-4 Regression coefficients for normalized release and pH using 28-d MCC-1 method .... 12-3012-5 Regression coefficients for normalized release and pH using the 7-d product
consistency test method .................................................. 12-31
xvi
ACRONYMS/ABBREVIATIONS
2D Two-dimensional3D Three-dimensionalASTM American Society for Testing and MaterialsBDAT Best demonstrated available technologyBO Bridging oxygenCNWRA Center for Nuclear Waste Regulatory AnalysesCVS Compositional variability studyDOE U.S. Department of EnergyDSC Differential scanning calorimetryDWPF Defense Waste Processing FacilityEA Environmental AssessmentEPA U.S. Environmental Protection AgencyESM Engineering-scale melterFEH Free energy of hydrationHLW High-level wasteHWVP Hanford Waste Vitrification PlantIDMS Integrated [Defense Waste Processing Facility (DWPF)] melter systemIFO Index of feed oxidationKfK Kernforschungszentrum KarlsruheLAW Low-activity wasteMCC Materials Characterization CenterNBO Nonbridging oxygenNCAW Neutralized current acid wasteNRC U.S. Nuclear Regulatory ComnmissionPCT Product consistency testPFA PerfluoroalkoxyPHA Precipitate hydrolysis aqueousPNNL Pacific Northwest National LaboratoryPUREX Plutonium Uranium Extraction ProcessR&D Research and developmentRSM Research scale melterSA/V Surface area/volumeSBS Structural bond strengthSGM Scale glass melterSNF Spent nuclear fuelSPFT Single-pass, flow-throughSRL Savannah River LaboratorySRS Savannah River SiteSVS Scaled Vitrification SystemTHOREX Thorium Extraction ProcessTOC Total Organic CarbonTTT Temperature-time-transformationTWRS Tank Waste Remediation System
xvii
ACRONYMS/ABBREVIATIONS (cont'd)
TWRS-P Tanks Waste Remediation System-PrivatizationVFT Vogel-Fulcher-TammannWAPS Waste acceptance product specificationsWP Waste packageWVDP West Valley Demonstration Project
xviii
ACKNOWLEDGMENTS
This report was prepared to document work performed by the Center for Nuclear Waste Regulatory
Analyses (CNWRA) for the U.S. Nuclear Regulatory Commission (NRC) under Contract No.
NRC-02-97-009. The activities reported here were performed on behalf of the NRC Office of Nuclear
Material Safetyand Safeguards, DivisionofFuel Cycle Safetyand Safeguards. Thereportis anindependent
product of the CNWRA and does not necessarily reflect the views or regulatory position of the NRC.
The authors wish to acknowledge L. Yang, R Pabalan, and N. Sridhar for technical reviews; W. Patrick for
programmatic review; B. Long, J. Pryor, C. Cudd, C. Gray, and A. Woods for editorial reviews; and the
assistance of J. Gonzalez, L. Selvey, A. Ramos, and J. Wike for preparation of the document.
QUALITY OF DATA, ANALYSES, AND CODE DEVELOPMENT
DATA: No CNWRA-generated original data are used in this report.
ANALYSES AND CODES: No computer codes were used for analyses contained in this report.
xix
EXECUTIVE SUMMARY
This report provides a review of the glass melt properties and characteristics, the basis and importance of
imposed constraints on the design of glass composition, and the potential safety and processing aspects of
glass melting that can compromise radiological safety. This report has been written to assist the U. S. Nuclear
Regulatory Commission (NRC) in
* Determining if sufficient information exists to assess safety considerations regarding vitrification of the
waste stored at the Hanford site near Richland, Washington* Determining whether current regulatory guidelines are adequate to control implementation of the
vitrification process to immobilize high-level waste (HLW) at Hanford* Identifying technical uncertainties* Assessing where future guidance may be warranted
Research and development (R&D) of waste forms for immobilization of liquid radioactive wastes began in
the mid-I 950s. Initial investigations included evaluation of waste forms such as borosilicate glass, phosphate
glass, nepheline-syenite glass, avarietyofpolyphase ceramics, and bituminous and concrete materials. The
R&D continued through the 1970s with no coherent approach. Most of the programs did not even
contemplate scaling up from laboratory studies to remote operations capable of handling millions of gallons
of waste. A systematic evaluation using standardized tests was initiated by the U.S. Department of Energy
in 1977 to select a waste form for HLW in the United States. This evaluation resulted in the selection of a
borosilicate glass-based waste form as a reference waste form for immobilization of HLW in the United
States. Currently, a vitrified waste form produced by melting HLW with glass forming oxides is being used
in the United States, France, England, Germany, Belgium, Japan, the Former Union of Soviet Socialists
Republic, and India as the preferred technology for the disposal of HLW. In the United States, HLW
vitrification is ongoing attheWestValleyDemonstrationProject(WVDP), WestValley, NewYork and the
Defense Waste Processing Facility (DWPF), Aiken, South Carolina, while vitrification is currently planned
for the HLW contained in 177 aging underground storage tanks at the Hanford site in the state of Washington.
The current plans for Hanford wastes call for pretreatment of wastes, prior to vitrification, that will separate
the waste into HLW and low-activity waste (LAW) fractions. The separated HLW and LAW fractions will
be vitrified into borosilicate glass.
The vitrificationtechnology, as well as guidelines for the disposal at aproposed geologic repository, imposes
stringentrequirements for producing an acceptable waste-form product. The glass composition designed for
vitrification should meet processing constraints, such as electrical conductivity (chapter 5), viscosity
(chapter 6), liquidus temperature (chapter 8), redox condition ofthe melt (chapter 9), solubility of noble metals
(chapter 10), and sulfur solubility limits (chapter 11); and product constraints such as glass transition
temperature (chapter7), chemical durability (chapter 12), and phase stability (chapter 13). Waste acceptance
criteria for the producers ofthe vitrified waste form are outlined inWaste Acceptance Product Specifications
xxi
(WAPS) by the U.S. Department of Energy (1996). The WAPS specify technical requirements that thewaste form must meet and the documentation the producers must provide to assure the product complies withthe NRC requirements in 10 CFR Part 60.1
The electrical conductivity in most oxide glasses is attributed to migration of positively charged alkali ions inthe presence ofan applied electric field. Joule-heated melters fornuclear waste vitrification require electricalconductivity of the glass melts between 10 and 100 S/m at melting temperature. The lower limit of 10 S/mfor electrical conductivity is imposed on the glass melts at melting temperature to ensure the electricalconductivity of the glass melt is significantly higher than the electrical conductivity of the refractoriessurrounding the glass melt, which ensures the current does not flow through the refractories. The upper limitof 100 S/m for electrical conductivityis imposed on the glass melt at meltingtemperature to ensure the glassmelt provides enoughjoule heating for melting without exceeding the maximum operating current density forAlloy 690 electrodes. The glass components, Li2O and Na2O, have the strongest tendencies to increaseelectrical conductivity, while SiO2 has the strongest tendency to decrease electrical conductivity. Chapter 5provides a review of published studies on electrical conductivity and its effect on the nuclear wastevitrification process.
Viscosity is a property of a liquid state and a measure of the resistance to shear deformation. Joule-heatedmelters for nuclear waste vitrification require the viscosity of glass melts to be between 2 and 10 Pa-s (20 and100 P) at melting temperature. The limits are imposed on glass melts at melting temperature to ensure theglass melt is fluid enough to homogenize and pour. If the viscosity of the melt is below 2 Pa-s (20 P), the glassmelt could increase erosion ofthe refractory, volatilization of alkalis, boron, and radionuclides; penetration ofmelt along the refractory joints; or settling of noble metals. If the viscosity is greater than 10 Pa-s (100 P),glass melt could plug the pour spout during pour, have undissolved components in the melt, or crystallize atcold spots in the melter. The glass components SiO2, A1203, and ZrO2 (in order) have the strongest effectsfor increasing the melt viscosity, while Li2O, Na2O, CaO, and B203 have the strongest effect for decreasingthe melt viscosity. Chapter 6.0 provides a review of published studies onviscosityand its effect onthenuclearwaste vitrification process.
At the time of shipment to a federal repository, the WAPS requires the producer certifythat, after the initialcool-down, the waste formtemperature has not exceeded 400 'C. This product specification was establishedto ensure the waste form is in a solid form at the time of shipment. The glass transition temperature (Tg)range marks the temperature interval over which a given system gradually transforms while cooling from asupercooled liquid state into the glass state. There is no absolute measure ofTg because its value is influencedbythethermal historyand rate of cooling. Minor variations in composition, redox, or cooling rate do not showsignificant change in T. Chapter 7 provides a review of published studies on glass transition.
Liquidus temperature (TL) is the highest temperature at which melt and primary crystalline phases can coexistat equilibrium. At temperatures higher than TL, no crystalline phases are present in the melt. If the lowesttemperature in the melter is lower than the TL of the melt, crystalline phases can precipitate and causeprocessing problems. If the crystalline phases are electrically conductive, electrical shorting could occur in
"The current regulation governing deep geologic disposal of high-level waste is 10 CFR Part 60. At the direction ofCongress, the U.S. Nuclear Regulatory Commission has proposed a regulation at 10 CFR Part 63 that would govern disposal at theproposed repository site at Yucca Mountain, Nevada.
xxii
the melter. In joule-heated melters operating at an average temperature of 1,150 0C, the TL limit is set at
1,050 'C. Chapter 8 provides a review of published studies on TL and its effect on the nuclear waste
vitrification process.
Redox (reduction-oxidation) reactions involve a transfer of one or more electrons within the different
oxidation states of a single multivalent element or from one multivalent element to another within a chemical
system. Multivalent elements that can coexist in multiple valance states are redox couples (e.g., Fe redox
couple can exist as either Fe2+/Fe3" or Fe0/Fe2+, depending on the available oxygen in the system). A glass
melt is defined as extremely reducing if the Fe2+/Fe3+ ratio is greater than one. Under such conditions,
sufficient accumulation of conductive metals and metal sulfides could occur and short-circuit the melter. If
the Fe2+/Fe3 ' ratio is less than 0. 01, a glass melt is defined as extremely oxidizing. Under extremely oxidizing
conditions, foaming is observed in the melter. Foam creates an insulating layer of gas bubbles between the
cold cap and the melt, disruptingthe thermal gradients in the melter. Knowledge of participating redox couples
in redox reactions in glass melts and its temperature and compositional dependence is of great importance
during vitrification of HLW in joule-heated melters as discussed in chapter 9.
Noble metals such as Ru, Rd, and Pd in the HLW originate from the fission of U-235. Noble metals in the
borosilicate glass melts have been a major concern in the HLW vitrification process because of their low
solubility, high volatilization rate, and high electrical conductivity. In 1985, the accumulation ofthe noble metals
on the floor of the Pamela melter, in Mol, Belgium, resulted in electrical shorting of ajoule-heated melter.
Chapter 10 provides a review of published studies on effects of the noble metals on the nuclear waste
vitrification process.
Borosilicate glass-based waste forms have limited sulfur solubility. In the WVDP and DWPF waste forms,
SO3 levels are maintained below 0.25 wt%to avoid formation ofimmiscible sodiumsulfate phaseinthe melt.
Traditionally, wastes containinghigh concentrations of sulfur are washed withwaterto remove soluble sulfate
salts. If the concentration of S03 is in excess of 0.25 wt%, the SO3 concentration limits the waste loading in
the glass. The SO3 solubility limit is established based on the operating redox range of 0.01 to 0.5. In
chapter 11, adiscussiononthe sulfur solubilitylimitimposed onthe glass composition is followed byareview
of various studies on sulfur solubility and its implication on the LAW vitrification at the Hanford site.
When glass comes in contact with the repository environment, such as flowing or stagnant groundwater, acid
rain, corrosive gases and vapors, or aqueous solutions, chemical reactions occur at the surface and spread
to the whole of the glass, depending on its composition, pH of the solution, and the temperature of the
environment. In chapter 12, areview of recent advances in understanding the dissolution behavior of glasses
as a function of composition, pH, and temperature is provided.
Phase stability in glass can be affected by either liquid-liquid phase separation or crystallization on cooling
from the melt. For the glass waste form, both the phase separation and crystallization processes can result
in the development of an inhomogeneous microstructure that may affect the reliability of the waste glass
process and product performance. The phase stability requirements, which include development of a
temperature-time-transformation (TTT) diagram for each projected waste type, have been outlined in the
WAPS. Crystallization can occur in nuclear waste glasses during continuous cooling of pour canisters and
annealing of quenched glasses. A TTT diagram is a useful tool for scoping the heat treatment conditions for
crystallization. Experimental results indicate that both liquid-liquid phase separation and crystallizationhave
a detrimental effect on the durability of nuclear waste glasses. Controlling the glass compositions and defining
xxiii
the heat treatment conditions that avoid phase separation and crystallization are keys to achieving waste glassdurability control during processing, as discussed in chapter 13.
In summary, the review ofthe history ofwaste form development, and the information (data and models) onglass-melt properties for various types of commercial and waste-glass compositions indicates that the behaviorof glasses is complex and cannot be expressed in simple terms for all wastes. The glass composition needsto be designed uniquely for each waste type. The lessons learned, or the existing database of HLW glasses,can be used as a guideline to narrow the scope of R&D needed to meet the requirements.
REFERENCE
U.S. Department of Energy. Waste Acceptance Product Specifications for Vitrified High-Level WasteForms. EM-WAPS. Revision 2. Washington, DC: U.S. Department of Energy. 1996.
Xxiv
1 INTRODUCTION
1.1 BACKGROUND
The tank waste remediation system (TWRS) program at the Hanford, Washington, site wasestablished in 1991 bythe U.S. Department of Energy(DOE)to manage the cleanup ofthenearly54 milliongal. (204,000 in3 ) of radioactive waste currentlyinunderground storage tanks. In addition, the DOE choseto use a privatization-type contract for the construction and operation of the TWRS facility.
The tank waste remediation systems privatization (TWRS-P) program was divided into two phases;Phase I was the proof-of-concept or demonstration phase and Phase II was the full-scale operations phase.Phase I was further divided into two phases, Part A and Part B (U.S. Department of Energy, 1998). Aftercompletion of Part A of Phase I in fiscal year (FY) 1998, the DOE decided to further subdivide Part B intotwo parts, Phase I Part B-i and Phase I Part B-2. With this newly revised contract, DOE awarded acontractor team led by BNFL Inc. to proceed to the next phase (Phase I/Part B-1) of the TWRS-P project.This phase was a 24-mo design period during which BNFL Inc. would address technology scaleup; regulatory,financial, and permitting issues; and the safety basis for operations. As Part B-i of Phase I nearedcompletion, the contractor team led by BNFL Inc. presented a new estimate in April 2000 for Part B-2 ofPhase I, which would include the construction and operation ofa facilityto process approximately 10 percentof the Hanford tank waste by mass and 20-25 percent by radioactivity. This new estimate was significantlyhigher than a prior estimate provided bythe BNFL Inc. team approximately 18 mo earlier and resulted in theDOE decision to hold anew competition to seek other contractors for the project. In addition, it is not likelyDOE will continue with a privatization-type contract, and, as a result, the U.S. Nuclear RegulatoryCommission (NRC) may no longer be involved in the licensing and regulation of Phase II of the project asoriginallyplanned inthe Memorandum of Understandingbetween DOE and NRC.' Consequently, the Centerfor Nuclear Waste Regulatory Analyses, which has been engaged by the NRC to assist in development ofregulatoryand technical guidance forTWRS-P projectfacilities, isnowplanningto close all current work andcease involvement in the project at the end of FY 2000.
1.2 SCOPE OF REPORT
This report provides a review of glass-melt properties and characteristics, basis and importance ofimposed constraints on the design of glass composition, and potential safety and processing aspects of glassmelting that can compromise radiological safety. This report was written to assist the NRC in
* Determining if sufficient information exists to assess safety considerations regarding vitrification ofthe waste stored at the Hanford site near Richland, Washington
* Determining if current regulatory guidelines are adequate to control implementation ofthe vitrificationprocess to immobilize high-level wastes (HLW) at Hanford
1Memorandum of Understanding between the U.S. Nuclear Regulatory Commission and the U.S. Department of Energy,January 29, 1997, Federal Register, V. 62, No. 52, 12861, March 18,1997.
1-1
* Identifying technical uncertainties
* Assessing where future guidance may be warranted
In chapter 2 ofthis report, a brief history of waste form development is provided, which is followedin chapter 3 by the Waste Acceptance Product Specifications (WAPS) imposed on the HLW vitrified wasteform in the United States. In chapter 4, glass formation theories are discussed. These theories provide thebackbone of many models used to predict glass properties.
The importance of electrical conductivity of glass melts, its temperature and compositionaldependence, and its impact onthe operations of electrically (joule) heated melters are discussed inchapterS.The joule-heated melters for nuclear waste vitrification require electrical conductivity of the glass meltsbetween 1 0 and 1 00 S/in at the melting temperature.
In chapter 6, the role of melt viscosity in the processing of glass is discussed. The joule-heatedmelters for nuclear waste vitrification require viscosity ofglass melts between 2 and 1 0 Pa-s (20 and 1 00 P)at melting temperature. The limits are imposed on glass melts at melting temperature to ensure the glass meltis fluid enoughto homogenize and pour. The models used for predicting melt viscosity fromglass compositionare also presented.
The WAPS require, at the time of shipment to a federal repository, the producer to certify that afterthe initial cool-down, the waste form temperature has not exceeded 400 'C. This product specification wasestablished to ensure the waste form is in a solid form at the time of shipment. The glass transitiontemperature (Td) range marks the temperature interval over which a given system gradually transforms oncooling from a supercooled liquid state into the glass state. There is no absolute measure of Tg because itsvalue is influenced bythe thermal history and rate of cooling. The importance ofTg is discussed in chapter 7.
Liquidus temperature (TL) is the highest temperature at which the melt and primary crystalline phasescan coexist at equilibrium. If the lowest temperature in the melter is lower than the TL ofthe melt crystallinephases can precipitate and cause processing problems. In joule-heated melters operating at an averagetemperature of 1,150 'C, the TL limit is set at 1,050 'C. Therefore, the understanding of TL and itscompositional dependence is important for the operations ofjoule-heated melters, and TL is discussed inchapter 8.
Redox (reduction-oxidation) reactions involve a transfer of one or more electrons within the differentoxidation states of a single multivalent element or from one multivalent element to another within a chemicalsystem. An example of a multivalent element that can coexist in multiple valance states is Fe, which can existas either Fe2 l/Fe& or Fe0/Fe2+depending on the available oxygen in the system. Knowledge ofredox reactionsinglass melts and their temperature and compositional dependence duringvitrificationofHLWinjoule-heatedmelters is important. In chapter 9, a brief discussion of redox limits is followed by a discussion on thethermodynamic basis for redox reactions and the variables that impact redox reactions in glass melts. A redoxresponse within a specified redox range is necessary to avoid process upsets that eventually could lead topermanent melter damage. A glass melt is defined as extremely reducing if the Fe2+/Fe3+ ratio is greaterthan 1. Under such conditions, sufficient accumulation of conductive metals and metal sulfides could occurand short-circuit the melter. If the Fe2+/Fe 3+ ratio is less than 0.01, a glass melt is defined as extremely
1-2
oxidizing. Under extremelyoxidizing conditions, foaming is observed in the melter. Foam creates aninsulating
layer of gas bubbles between the cold cap and the melt, disrupting the thermal gradients in the melter.
Noble metals such as Ru, Rh, and Pd inthe HLW originate from the fission of U-235. Noble metalsin the borosilicate glass melts have been a major concern in the HLW vitrification process because of their
low solubility, high volatilization rate and high electrical conductivity. In silicate glasses, solubilities of Rh, Pd,
and Ru are approximately 0.05, 0.03, and 0.01 wt%, respectively. In 1985, the accumulation of the noblemetals onthe floor ofthe Pamelamelter, in Mol, Belgium, resulted in electrical shorting ofthejoule-heatedmelter. Potential issues related to noble metals in melts are discussed in chapter 10.
Borosilicate glass-based waste forms have limited sulfur solubility. In the West Valley DemonstrationProject (WVDP) and the Defense Waste Processing Facility (DWPF) waste forms, S0 3 levels are
maintained below 0.25 wt%to avoid formationofimmisciblesodiumsulfate phase inthe melt. Traditionally,wastes containing high concentrations of sulfur are washed to remove soluble sulfate salts. If the
concentration of S0 3 is in excess of 0.25 wt%, the S0 3 concentration limits the waste loading in the glass.The S03 solubility limit is established based on the operating redox range of 0.01 to 0.5. In chapter 11, a
discussion of sulfur solubilitylimits imposed onthe glass composition is followed byareview of various studieson sulfur solubility and the implication on the low-activity waste (LAW) vitrification at the Hanford site.
When glass comes in contact with the repository environment, such as flowing or stagnant
groundwater, corrosive gases and vapors, or aqueous solutions, chemical reactions occur at the surface that
spread to the whole ofthe glass depending on its composition, the pH ofthe solution, and the temperature of
the environment. In chapter 12, a review of recent advances in the understanding ofthe dissolution behaviorof glass as a function of glass composition, pH, and temperature is provided.
Crystallization of glass during cooling results inthe formation of several phases in the glassy matrix.
These phases can affect glass durability or even the glass melting behavior. In chapter 13, a review of thecrystallization behavior of various waste forms is provided.
1-3
2 HISTORY OF WASTE FORM RESEARCH, EVALUATION,AND SELECTION
Research and development (R&D) of waste forms to immobilize liquid radioactive wastes began in the mid-
1950s. Initial investigations included evaluating waste forms, such as borosilicate glass, phosphate glass,nepheline-syenite glass, a variety of polyphase ceramics, and bituminous and concrete materials (Hench etal., 1984). The R&D continued through the 1970s withno coherent approach. Most ofthe programs did noteven consider scaling up from laboratory studies to remote operations capable ofhandling millions of gallonsof waste. According to Hench (1995), efforts to compare various waste forms uncovered a potentialproblem-lack of standardized test methods to evaluate waste form materials. In 1977, the DOE initiated acomprehensive program to select a waste form for HLW in the United States based on standardized tests.The waste forms were reviewed againin 1994 forthe Puimmobilization program. In addition, several review
papers were published to summarize and compare various waste forms [e.g., Donald et al. (1997) providesthe most current status of waste form for immobilization].
2.1 HIGH-LEVEL WASTE TECHNOLOGY PROGRAM
In 1979, the DOE initiated the National High-Level Waste Technology Program, which sponsoredR&D at 14 laboratories, 3 universities, 3 industrial laboratories, and several DOE sites. The study evaluated17 different waste forms as shown in table 2-1. In 1980, R&D activities for 10 waste forms were terminatedbased on reviews that indicated significant technical concerns about their viability as candidates for disposal(Hench et al, 1984). The remaining seven waste forms underwent the following assessments by the DOEteams:
* Evaluations at the DOE defense waste sites* Peer-review evaluation* Product performance evaluation* Processability analysis
Based on the combined results, borosilicate glass was selected as the reference, and Synroc wasselected as an alternative waste form for the U.S. HLW program. Important attributes supporting theselection of borosilicate glass as the reference waste form were
* Ability to incorporate a wide range of waste compositions* Good durability in potential repository environments* Low processing temperature* Remote processing* Relative insensitivity to radiation damage
Processability of borosilicate glass had an advantage factor of 2 to 4 times over Synroc. Thisadvantage was attributed to the successful operation of calcine-fed and slurry-fed melters at Savannah RiverLaboratory (SRL) and proven operations at the Atelier de Vitrification de Marcoule facility in France.However, the relative economics of glass compared to Synroc was under debate because Synroc offered
three times higher waste loading compared to borosilicate glass, which offered potential economic advantage.
2-1
Table 2-1. Candidate waste forms considered for geologic disposal of high-level waste (Hench etal., 1984)
Waste form Comments
Borosilicate glass Reference waste form
Synroc-C, -D* Alternative waste form
Tailored ceramic Semifinalist alternative waste form
High-silica glass Semifinalist alternative waste form
Formed under elevated temperature and pressure Semifinalist alternative waste formconcrete
Coated sol-gel particles Semifinalist alternative waste form
Glass marbles in Pb matrix Semifinalist alternative waste form
Phosphate glass Eliminated in 1979-1980
Clay ceramic Eliminated in 1979-1980
Titanate ion-exchanger Eliminated in 1979-1980
Stabilized calcine Eliminated in 1979-1980
Pelletized calcine Eliminated in 1979-1980
Normal concrete Eliminated in 1979-1980
Hot-pressed concrete Eliminated in 1979-1980
Cermett Eliminated in 1979-1980
Disc-pelletized coated particles Eliminated in 1979-1980
Synroc-C composition was originally proposed for immobilization of commercial reactor wastes, while Synroc-D was amodification of Synroc-C developed for immobilization of defense wastes.tCermet fixes high-level waste in a continuous, corrosion-resistant, thermally conductive, metal-alloy matrix
2-2
Vitrification, using borosilicate glass waste forms, is successfully used in the United States, United
Kingdom, France, Germany, Belgium, the Former Union of Soviet Socialists Republic, India, and Japan for
immobilizing HLWs, while DOE-sponsored work on Synroc was discontinued effective January 1983. A
review by Jain (2000) on vitrification facilities worldwide provides the current status, process description,lessons learned, and potential safety issues at various vitrification facilities.
2.2 PLUTONIUM IMMOBILIZATION PROGRAM
The waste forms for immobilization were revisited under the DOE Office of Fissile MaterialsDisposition program created in 1994 to manage DOE activities relating to the management, storage, and
disposition of surplus fissile materials. The goal was to make Pu as unattractive and inaccessible for retrieval
and use in weapons as the residual Pu in spent fuel from commercial reactors (Bronson et al., 1998). Under
this program, based on the literature review, 70 waste forms were identified for immobilizing radioactivewastes. Initial screening was conducted based on pass/fail criteria shown in table 2-2. Failure of any one
Table 2-2. Initial screening criteria for Pu immobilization waste forms (Lawrence LivermoreNational Laboratory, 1996)
Criteria Requirement J Basis l
No free water "Shall not contain free liquids in amount that 10 CFR 60.1 35(b)(2)could compromise the waste package."
Solidification and "Shall be ... in solid form ... [and] ... 10 CFR 60.135(c)(1&2)
consolidation consolidated ... to limit the availability andgeneration of particulate."
Stability "Shall not contain explosive or pyrophoric or 10 CFR 60.135(b)(1)&(c)(3)chemically reactive materials in an amountthat could compromise the waste package..."and "...shall be noncombustible ... "
Criticality control "Keff must ... show at least 5 % margin" 10 CFR 60.131(b)(7)
Resource Conservation Cannot contain significant quantities of the 40 CFR 261.24Recovery Act metal following free metals: arsenic, barium,content cadmium, lead, mercury, selenium, and silver.
Readiness Must be technically viable for use 20 yr after U.S. Department of Energythe record of decision (subsequently, 10 yr). prescribed requirement.
Loading Must maintain a feasible volume of waste for Reasonable limit, factoring instorage increased resources required to
handle excess volumes ofimmobilized material andlimited storage sites/volumes.
2-3
criterion or "probable" failure of two criteria resulted in elimination of the waste form. Based on thisevaluation, 16 waste forms shown in table 2-3 were selected for stage 2 evaluation.
The following attributes were used for the second stage of the screening process.
0
S
0
0
0
Resistance to Pu theft, diversion, and recovery by a terrorist organization or rogue nationTechnical viability, including technical maturity, development risk, and acceptability for repositorydisposalEnvironmental safety, and health factorsCost effectivenessTimeliness
Table 2-3. Final ranking of waste forms for Pu immobilization (Lawrence Livermore NationalLaboratory, 1996)
Waste form Utility
Borosilicate glass 0.89
Synroc 0.66
Phosphate glass 0.55
Monazite 0.49
Metallic alloy 0.47
High-silica glass 0.44
Formed under elevated temperature and pressure concrete 0.40
Hot-pressed concrete 0.24
Phosphate-bonded ceramic 0.17
Silica-zirc phosphate 0. 17
Ceramics in concrete 0.14
Iron-enriched basalt 0.13
Ceramic pellets in metal matrix 0.13
Supercalcine 0.08
Glass-ceramic monoliths 0.03
Cermet 0.00
*Cermet fixes high-level waste in a continuous, conosion-resistant, thennally conductive, metal-alloy matrix.
2-4
Based on the second stage screening, a well-balanced portfolio of waste forms consisting of
borosilicate glass, ceramic (Synroc), and metallic alloywas selected for further evaluation, while glass-bonded
zeolite was retained as a reasonable alternative. Borosilicate glass was chosen because it is consistently the
highest performing waste form over all other glasses. Ceramic waste form was selected because it performedbetter than all other glasses except borosilicate glass. Metallic alloy was selected over concrete formed under
elevated temperature and pressure. However, for the disposal of Al-based spent nuclear fuel, in 2000 DOE
selected melt dilutetechnologyas a preferred alternative. The melt-dilute technologyproduces ametallic-alloywaste form (Adams et al., 2000).
In September 1997, based on the technical evaluation and assessments, Synroc was recommendedfor deployment inthe planned Pu immobilization plant based on the following advantages over borosilicateglass.
Ceramic waste form is expected to be much more durable in the repository environment and
should retain Pu and its decay products for longer periods than the glass waste form.
* Ceramic waste form process has a significantly lower source term and potential for workerexposure, which enhances its preference by reason of the as-low-as-is-reasonably-achievablestandard.
* Ceramic waste form and process offer significant cost savings compared to the glass technology.
* Ceramic waste form is somewhat more robust to the threat of theft and diversion by terroristsor rogue states.
* Maturity of the ceramic waste form technology was found to be comparable to glass.
2.3 SELECTION OF A WASTE FORM FOR HANFORD HIGH-LEVEL WASTE
Westinghouse Hanford Company (1990) wrote a white paper tojustify its selection of borosilicateglass as a suitable waste form for the HLW stored at the Hanford site. The paper cites the followingstrengths of the borosilicate glass waste form.
* Based on independent investigation conducted in many countries, borosilicate glass waste form
is the dominant choice for immobilization.
* Thirteen borosilicate glass-based vitrification facilities are in operation or under constructionaround the world.
* Decisions to convertHLW into borosilicate glass were made atWVDP and DWPF in 1982, andfor Hanford wastes in 1988 but only after an in-depth examination of the alternatives.
* Borosilicate glass waste form is capable of meeting all waste form specifications, includingNRCand U.S. Environmental Protection Agency (EPA) regulations for geologic disposal of HLW.
2-5
* Because of the wide acceptance of borosilicate glass as a waste form, there has been relativelylittle investigation of alternative waste forms.
* In 1990, EPA promulgated vitrification as the treatment standard [i.e., best demonstratedavailable technology (BDAT)] for the high-level fraction of the mixed waste generated duringreprocessing of nuclear fuel. In the ruling, EPA concluded that vitrification will provide effectiveimmobilization ofthe inorganic Resource Conservation and Recovery Act hazardous constituentsin the high-level mixed waste generated during the reprocessing of fuel rods.
Vitrification is currently planned for the HLW contained in 177 aging underground storage tanks atthe Hanford site. The current plans for Hanford wastes call for pretreatment of the wastes prior tovitrification that will separate the waste into HLW and LAW fractions. The HLW and LAW will beseparatelyvitrified into borosilicate glass. The status ofthe wastes stored at the Hanford site and its chemistryhave been reviewed by Cragnolino et al. (1997) and Pabalan et al. (1999).
2.4 CURRENT STATUS OF VARIOUS WASTE FORMS
Review of the various waste forms indicates that borosilicate glass is the generally accepted wasteform for the immobilization ofHLW and low-level waste. The borosilicate glass waste form-based vitrificationtechnologyis mature, proven, and safe, but with the emergence ofnew sources of radioactive materials, suchas surplus Pu, requiring immobilization, there is renewed interestinthe ceramic (Synroc) waste form. Donaldet al. (1997) provided a comprehensive review of various waste forms for radioactive wastes. This referenceprovides the most current status of suitability ofvarious waste forms for immobilization. Donald et al. (1997)state that although, atthe present time, borosilicate glass is the generally accepted first generation waste formfor the immobilization of HLW, the emergence of additional sources ofhighlyradioactive materials requiringimmobilization has renewed the interest in a reexamination of many of the alternative candidates such asSynroc-based titanate, zirconate, and phosphate ceramics, iron phosphate glasses, and basaltic glass-ceramics.Except for Synroc, the majority of alternate waste forms have been studied only marginally.
2.5 REFERENCES
Adams, T.M., H.B. Peacock, N.C. Iyer, M.W. Barlow, H.M. Brooks, W.F. Swift, and F.C. Rhode.Melt-Dilute Treatment Technology for Aluminum-Based Research Reactor Fuel.WSRC-MS-2000-00206. Aiken, SC: Westinghouse Savannah River Company. 2000.
Bronson, M., J. Diaz, T.A. Edmunds, L.W. Gray, D.C. Riley, and T.L. Rising. Plutonium immobilization formevaluation. Proceedings of the Waste Management '98 Symposium. CD Rom version: PaperNo. 65-02. Tucson, AZ: WM Symposia Inc. 1998.
Cragnolino, G.A., M.S. Jarzemba, J. Ledbetter-Ferrill, W.M. Murphy, R.T. Pabalan, D.A. Pickett,J.D. Prikryl, and N. Sridhar. Hanford Tank Waste Remediation System Familiarization Report.CNWRA 97-001. San Antonio, TX: Center for Nuclear Waste Regulatory Analyses. 1997.
Donald, I.W., B. L. Metcalfe, and R.N.J. Taylor. Reviewthe immobilization of high-levelradioactive wastesusing ceramics and glasses. Journal of Materials Science 32: 5,851-5,887. 1997.
2-6
Hench, L.L., D.E. Clark, and J. Campbell. High-level waste immobilization forms. Nuclear and ChemicalWaste Management 5: 149-173. 1984.
Hench, L.L. The '70's: From selection of alternative waste forms to evaluation of storage system variables.Proceedings of the International Symposium on Environmental Issues and Waste ManagementTechnologies in the Ceramic and Nuclear Industries. 97 h Annual Meeting of the AmericanCeramic Society, Cincinnati, Ohio, May 1-5, 1995. Ceramic Transactions Volume 61.Westerville, OH: American Ceramic Society 129-137. 1995.
Jain, V. Survey of Waste Solidification Process Technologies. NUREG/CR-6666. Washington, DC: U.S.Nuclear Regulatory Commission. 2000.
Lawrence Livermore National Laboratory. Fissile Material Disposition Program Screening ofAlternate
Immobilization Candidates for Disposition of Surplus Fissile Materials. UCRL-ID-1 18819,L-207790-1. Livermore, CA: Lawrence Livermore National Laboratory. 1996.
Pabalan, R.T., M.S. Jarzemba, D.A. Pickett, N. Sridhar, J. Weldy, C.S. Brazel, J.T. Persyn, D.S. Moulton,J.P. Hsu, J. Erwin, T.A. Abrajano, Jr., and B. Li. Hanford Tank Waste Remediation System
High-Level Waste Chemistry Manual. NUREG/CR-5751. Washington, DC: U.S. NuclearRegulatory Commission. 1999.
Westinghouse Hanford Company. Evaluation and Selection of Borosilicate Glass as the WasteformforHanfordHigh-LevelRadioactive Waste. DOE/RL-90-27. Richland, WA: Westinghouse HanfordCompany. 1990.
2-7
3 WASTE ACCEPTANCE CRITERIA FOR GEOLOGIC DISPOSAL
Waste acceptance criteria for the producers of the vitrified waste form are outlined in the WAPS documentby the DOE (1996). WAPS define the characteristics of the vitrified product and the canistered waste form
that will be produced at the vitrification plants and are the first step in the waste acceptance process. WAPSspecify technical requirements that the waste form must meet and the documentation the producers mustprovide to assure that products comply withthe NRC requirements in 10 CFRPart 60' (Stahl and Cloninger,1991). The WAPS are shown in table 3-1.
3.1 PRINCIPAL WASTE ACCEPTANCE CRITERIA
Since the intent of this report is to present the chemistry of the vitrification process and to focus onprocessing and safety issues that could result from inadequate control during processing, the discussion is
limited to waste form properties and characteristics that are important in producing waste forms. To meet
waste form specifications, the glass composition of the vitrified waste form must be designed to ensure it
meets WAPS and is also safely processible usingjoule-heating melter technology. WAPS that are dictatedby the glass composition and vitrification process are
* Chemical composition. The WAPS state that the "...producer shall project the chemicalcomposition, identify crystalline phases expected to be present, and project the amount of eachcrystalline phase, for each waste type," and the "...producer shall report the oxide compositionofthe waste form. The reported composition shall include all elements, excluding oxygen, presentin concentrations greater than 0.5 percent by weight of the glass, for each waste type."
* Product Consistency. The WAPS state that "the producer shall demonstrate control of wasteform production by comparing, either directly or indirectly, production samples to theEnvironmental Assessment (EA) benchmark glass. The consistency of the waste form shall bedemonstrated using the Product Consistency Test (PCT). For acceptance, the meanconcentrations of lithium, sodium, and boron in the leachate, after normalizing for theconcentrations in the glass, shall each be less than those of the benchmark glass... ." "Oneacceptable method of demonstrating thatthe acceptance criteria are met, would be to ensure thatthe mean PCT results for each waste type are at least two standard deviations below the meanPCT of the EA glass."
* Phase Stability. The WAPS state that, "the producer shall provide the following data for eachprojected waste type: (a) The glass transition temperature and (b) time-temperature-transformation diagram that identified the duration of exposure at any temperature that causessignificant changes in either the phase structure or the phase compositions....Producer shallcertify that after the initial cool-down, the waste form temperature has not exceeded 400 C."
'The current regulation governing deep geologic disposal of high-level waste is 10 CFR Part 60. At the direction of
Congress, the U.S. Nuclear Regulatory Commission has proposed a regulation at 10 CFR Part 63 that would govern disposal at the
proposed repository site at Yucca Mountain, Nevada.
3-1
Table 3-1. Waste acceptance preliminary specifications (U.S. Department of Energy, 1996)
[ Area Specification
Waste form specification 1.1 Chemical specification
1.2 Radionuclide inventory specification
1.3 Specification for product consistency
1.4 Specification for phase stability
1.5 Hazardous waste specification
Canister specifications 2.1 Material specification
2.2 Fabrication and closure specification
2.3 Identification and labeling specification
2.4 Specification for canister length and diameter
Canistered waste form specifications 3.1 Free-liquid specification
3.2 Gas specification
3.3 Specification for explosiveness, pyrophoricity, andcombustibility
3.4 Organic materials specification
3.5 Chemical compatibility specification
3.6 Fill height specification
3.7 Specification for removal of radioactive
3.8 Heat generation specification
3.9 Specification for maximum dose rates
3.10 Subcriticality specification
3.11 Specification for weight, length, diameter, andoverall dimensions
3.12 Drop test specification
3.13 Handling features specification
Quality assurance 4.0 Quality assurance specifications
Documentation and other requirements 5.1-5.14 (14 different specifications)
3-2
A vitrified waste form can meet waste form specifications if the glass melt properties such as viscosity,
electrical conductivity, TL, Tg, and melt characteristics such as redox, volatilization, phase separation, andprecipitation are defined and controlled during processing. Both product specifications in the WAPS andprocess requirements determined based on technology have to be considered together in developing glasscomposition for vitrified waste form. In this report, the role of various glass properties and characteristicsin optimizing the glass melt chemistry and chemical composition to safely produce a vitrified waste form isdiscussed.
3.2 REFERENCES
Stahl, D., and M.O. Cloninger. Waste acceptance for vitrified high-level waste forms. Proceedings of theFifth International Symposium on Nuclear Waste Management IV 9 3 rd Annual Meeting of theAmerican Ceramic Society, Cincinnati, Ohio, April 29-May 3, 1991. Ceramic TransactionsVolume 23. G.C. Wicks, D.F. Bickford, and L.R. Bunnell, eds. Westerville, OH: American CeramicSociety: 433-441. 1991.
U.S. Department of Energy. Waste Acceptance Product Specifications for Vitrified High-Level WasteForms. EM-WAPS. Revision 2. Washington, DC: U.S. Department of Energy. 1996.
3-3
4 GLASS FORMATION
The word glass is derived from a Latin term glaesum, which refers to a lustrous or transparent material.
Glass is also referred to as vitreous, originating from the Latin word vitrum. The American Society for
Testing and Materials (ASTM) defines glass as an inorganic product of fusion that has been cooled to a rigid
condition without crystallizing. This ASTM definition is fairlyrestrictive, since several organic systems can
make glass, and processes such as sol-gel and chemical vapor deposition do not require fusion to make glass.
To avoid such restrictions, glass is defined here as solid with a liquid-like structure, anoncrystalline solid or
an amorphous solid. A typical volume-temperature relationship for a glass-forming liquid is shown in
figure 4-1. Glass-forming liquids when cooled from location a, skip the crystallizationtemperature atlocation
b and continue to follow the path shown. The glass-transition range starts at a point where the cooling path
begins to depart from the supercooled liquid path and ends when the cooling profile attains a constant slope
as shown in figure 4-1. This constant slope after the transition range is referred to as glassy state. Also
shown in figure 4-1 is the effect of the cooling rate. The faster the glass is cooled, the higher the specific
volume of glass.
co
0
E
a)E
C.)0)ft0na),
Temperature -
Figure 4-1. Specific volume-temperature diagram for a glass-forming liquid
4-1
4.1 GLASS FORMATION PRINCIPLES
Several structural theories-Goldschmidt's radius ratio criterion, Zachariasen's random networktheory, Smekal's mixed bonding rule, Stanworth's electronegativityrule, Sun's single-bond strength criterion,Dietzel's field strength criterion, and Phillips' topological constraint model-and kinetic theoryhave beenput forward to explain the formation of glass. These theories are discussed in detail in textbooks byVarshneya (1994), Paul (1 982), and Rawson (1967). In this chapter, Zachariasen's randomnetworktheoryand Sun's bond strength criterion, both of which had significant influences on our current understanding ofglass formation principles and are used byseveral researchers to model nuclear waste glasses, are reviewed.The kinetic theory, which is important in analyzing phase separation and crystallization in glass, is presentedin chapter 13.
4.1.1 Zachariasen's Random Network Theory
The random network theory was proposed by Zachariasen in 1932. Zachariasen argued that theinteratomic forces in glass must be similar to those in the corresponding crystals, since the mechanicalproperties of the two forms are similar. Similar to crystals, the atoms in the glass must form extendedthree-dimensional (3D) networks but not in a symmetrical or periodic pattern. He defined glass as a"substance (that) can form extended 3D networks lacking periodicity with an energy content comparablewith that of the corresponding crystal network" and laid the following four rules for glass formation in acompound AmOn:
(1) An oxygen atom is linked to no more than two atoms of A.(2) The oxygen coordination around A is small, say three or four.(3) The cation polyhedra shares corners, not edges or faces.(4) At least three corners are shared.
Based on the theory, Group I and II (oxides of formula A20 and AO) cations cannot satisfy rules 1to 3, and, therefore, do not form glass. For Group III oxides (A 203 ), rules 1, 3, and 4 are satisfied if theoxygens form triangles around A. B203 in this categoryis a glass former. Group IV oxides (AO2) and GroupV oxides (A 20 5) would form glass if the oxygen formed tetrahedra around the cations. Glass formers in thiscategory include Si0 2, GeO2, P205 , and As205. Oxides in Group VI, VII, and VIII can satisfy rules 1,3, and4 if cation octahedra are formed, but rule 2 is violated. Zachariasen's theory received wide support andacceptance with the following limitations (Paul, 1982).
* In most of the oxides, oxygenhas a coordinationnumber oftwo. However, inthe binaryT120-B203 system with low T120 concentration, the coordination number for oxygen may be three.
* The coordination number for Si, P, and B in glass are four, and three or four. However, thecoordination number of Te in the PbO-TeO2 glass system is six.
* Alkali phosphate glasses containing more than 50 mol% alkali oxide contain two-dimensional(2D) chains of various sizes. Thus, a 3D network may not be necessary.
4-2
4.1.2 Sun's Single-Bond Strength Criterion
In 1947, Sun proposed a single-bond strength criterion for glass formation. According to Sun, theprocess of glass formationinvolves inabilityofthe bonds inthe liquid state to rearrange during crystallizationand, hence, thehigherthe bond strengththe better the glass former. Bond strengths associated with variousoxides are shown in table 4-1. Based on the bond strengths, Sun classified the glass forming oxides asfollows:
* Oxides exceeding bond strength 334 kJ/mol (80 kcallmol) as glass network formers.* Oxides with bond strengths lass than 251 kJ/mol (60 kcallmol) as glass network modifiers.* Oxides with bond strengths between 251 and 334 kJ/mol (60 and 80 kcal/mol) as intermediates.
The important contribution made by Sun was the introduction of an intermediate category. Sun's
model is widely used in the development of glass property models based on a thermodynamic approach.
4.2 GLASS FORMING SYSTEMS
Tables 4-2 and 4-3 show the composition of several commercial glass-forming systems and nuclearwaste glasses, respectively. Silica based glass forming systems are widely used in commercial industry andconsist ofsilica as amajor constituent along with alkalis, alkaline earths, boric oxide, and lead oxide. To helpunderstand the behavior of nuclear-waste glasses and models used to relate glass properties to composition,a review of the basic glass-forming systems of nuclear waste glasses is provided in the following sections.
4.2.1 Vitreous Silica Glass
Silica glass is highly refractory and has high chemical resistance to corrosion, low electrical
conductivity, near-zero thermal expansion, and good ultraviolettransparency. Silica glass is widelyused inoptical fibers and astronomical mirrors. The structure of silica glass consists of slightly distorted SiO4
tetrahedrajoined to each other at the corners as shown in figure 4-2. Each oxygen acts as abridge betweenneighboring tetrahedra and is called bridging oxygen (BO). In borosilicate-based nuclear waste glasses, SiO4
tetrahedras are the basic building blocks of the glass structure.
4.2.2 Alkali Silicate Glass
Alkalis, as shown by Sun's bond strength criterion, are classified as network modifiers. When an
alkali (M) is added to silica glass, it enters the glass structure by breaking the -Si-O-Si- bond and attachingitself to the broken oxygen bond as shown by Eq. (4-1).
=_ Si - 0 - S I-= + M 2 0 A_ Si - O M+ + M+ O0 - Si _(4-1)
As shown byEq. (4-1), each alkali ion results inthe creation of one nonbridging oxygen (NBO). The
addition of alkali ions results inthe breakdown ofthe silica tetrahedra connectivity, and the glass propertiesare significantlyaffected. The effect of alkali ions on glass properties is presented in various chapters of thisreport.
4-3
Table 4-1. Single-bond strengths for oxides (Varshneya, 1994)
M in MO, Valence Dissociation Energy Ed. Coordination Single-BondI~ ~ ~ ~~MnO I Iaec per MO,, U(kcal) |Number |Strength W (kcal)!
Glass formers B 3 1.49 x 103 (356) 3 498 (119)Si 4 1.77 x I03 (44) 4 444 (106)Ge 4 1.80 x 103 (431) 4 452 (108)Al 3 1.68 x10-1.32x 103(402-317) 4 423-331(101-79)B 3 1.49 x 103(356) 4 373 (89)P 5 1.85x 103(442) 4 465-368(111-88)V 5 1.88 x 103 (449) 4 469-377(112-90)As 5 1.46 x 103 (349) 4 364-293 (87-70)Sb 5 1.42 x 103 (339) 4 356-285 (85-68)Zr 4 2.03 x 103 (485) 6 339(81)
Intermediates Ti 4 1.78 x 103 (435) 6 306 (73)Zn 2 0.60 x 103 (144) 2 301 (72)Pb 2 0.61 x 103 (145) 2 306 (73)Al 3 1.32 x 103-1.68 x 103 (317-402) 6 222-281 (53-67)Th 4 2.16 x 103(516) 8 268(64)Be 2 1.05 x 103 (250) 4 264 (63)Zr 4 2.03 x 103(485) 8 255 (61)Cd 2 0.50 x 103(119) 2 251 (60)
Modifiers Sc 3 1.52 x 103 (362) 6 251 (60)La 3 1.70 x 103 (406) 7 243 (58)Y 3 1.67 x 103 (399) 8 209 (50)Sn 4 1.16 x 103 (278) 6 193 (46)Ga 3 1.12 x 103 (267) 6 188 (45)In 3 1.08 x 103 (259) 6 180 (43)Th 4 2.16 X103 (516) 12 180 (43)Pb 4 9.71 x 102(232) 6 163 (39)Mg 2 9.30 x 102(222) 6 155 (37)Li 1 6.03 x 102 (144) 4 151 (36)Pb 2 6.07 x 102(145) 4 151 (36)Zn 2 6.03 x 102 (144) 4 151 (36)Ba 2 1.09 x 103(260) 8 138 (33)Ca 2 1.07 x 103 (257) 8 134 (32)Sr 2 1.07 x 103 (256) 8 134 (32)Cd 2 4.98 x 102 (119) 4 126 (30)Na 1 5.02 x 102 (120) 6 83.7 (20)Cd 2 4.98 x 102(119) 6 83.7 (20)K 1 4.82 x 102 (115) 9 54.4 (13)Rb 1 4.82 x 102 (115) 10 50.2 (12)Hg 2 2.85 x 102(68) 6 46.1 (11)Cs 1 4.77x 102(114) 12 41.9 (10)
4-4
M M- M M I
Table 4-2. Commercial glass compositions in oxides (wtv/o) (Varshneya, 1994)
1 Bettle I Lime I Borosi- Lead Glue
Vitloua Tonr Babu- Pyrex | Thermo- licat Table. Halogen Optic l
Silica Vycor late Window Continer Bdb Tubing wav Type meter Crown Lamp Lamp E Gi S Glss Flint
SiO: 10000 940 72.7 72.0 74.0 73.6 72.1 74.0 81.0 72.9 69.6 67.0 60.0 52.9 65.0 49.8
.41.0, -_ _ 0.5 0.6 1.0 1.0 1.6 0.5 2.0 6.2 - 0.4 14.3 14.5 25.0 0.1
BO, | 5.0 - - - - - - 120 104 9.9 - 9.2 - -
SO, -_ 0.5 0.7 Tr. - 0.3 - -
CaO - - 130 100 54 5.2 5.6 75 - 04 - 6.5 17.4 - -
MgO - 2.5 3.7 3,6 3.4 - -0.2 - -- 4.4 10. -
BaO | _ - - Tr. _ -- - 2.5 - 18.3 - - 13.4
PbO - - - - - - - 17.0 - - - 18.7
4 NaO | 1.0 13.2 14.2 15.3 16.0 16.3 18.0 45 9.8 8.4 6.0 0.01 - - 1.2
L-A40 - - - - 0.6 0.6 1.0 - - 0.1 8.4 9.6 Tr. 1.0 - 8.2
ZnO - _ - - - _ - - - - - - - - 8.0
.4-0, - - Tr. Tr. Tr. Tr. _ Tr. _ Tr. 0.3 Tr. - - - - 04
U 0
Table 4-3. Compositions of selected candidate waste glasses (mol%) (Ellison et al., 1994)
Glass SiO J B2 03 L12 0 A1 203 FeO23 CaO MgO Tb 2 _ I O - ZO 2 BaO _ Mo JHW-39-1 56.74 9.16 11.15 8.45 2.80 4.62 3.44 1.32 - - 0.32 0.04 0.15 0.14
SRL 131A 51.15 9.73 13.69 7.05 2.25 5.57 116 2.28 0.57 0.02 0.13 0.07 1.96 -
SRL 165A 58.09 6.41 11.56 9.24 2.64 4.85 1.91 1.15 0.12 0.03 0.35 0.03 2.60 -
SRL 202A 55.15 7.75 9.74 9.58 2.55 4.84 1.45 2.22 0.77 0.02 0.05 0.10 1.72 0.02
SRL 211 48.30 7.68 16.71 6.98 2.44 6.13 6.22 - - - 0.15 3.02 -
|WVCM-47 51.30 12.56 10.83 4.92 5.71 5.53 0.77 2.39 0.90 - 0.17 0.02 1.09 0.01
WVCM-50 48.38 12.94 11.59 5.45 7.10 5.48 1 .07 1 .44 0.77 0.03 0.24 0.09 1.02 0.02
WVCM-70, Ref-6 48.62 13.20 9.20 8.85 4.19 5.37 0.61 1.57 0.71 0.02 0.76 0.07 0.83 0.02
lWV205 51.75 9.84 12.20 7.13 2.23 5.08 0.75 2.22 0.86 - 1.73 0.27 1.34 -
GPNL 76-68 52.59 9.03 17.87 - 0.33 4.99 2 83 - 2.82 4.32 0.99 0.25 0.03 0.97
GP 98/12 57.98 9.39 16.79 - 0.93 0.04 4.75 4.88 2.70 - 0.53 0.12 0.16 0.40
UK 209 53.95 10.18 8.53 8.51 3.19 1.09 - 10.02 - 0.31 0.74 0.16 - 0.78
UK 189 44.78 20.36 8.03 8.00 3.20 1.09 - 10.02 - 0.32 0.73 0.16 - 0.78
C 31/3 46.84 4.82 4.34 2.66 8.17 0.75 5.55 2.89 2.84 4.86 2.71 7.82 0.32 1.33
B 1/3 37.80 7.46 4.97 6.51 10.18 0.77 5.79 2.41 4.63 3.59 2.14 8.20 0.32 1.33
VG 98/3 49.10 10.61 25.31 - 0.83 0.31 2.92 0.70 3.11 - 1.34 0.29 0.32 1.06
SON 58 55.86 21.01 11.68 - 0.76 0.29 - - - 1.96 0.50 - 0.16
SON 64 56.76 19.13 14.62 - 1.20 2.46 - - - 1.15 0.29 0.95
Table 4-3. Compositions of selected candidate waste glasses (mol%) (Ellison et aL, 1994) (cont'd)
Glass Sio 20 Na1 O L-0 AO. Fe CaO MgO I Izo Zro2 ( BaO I _0 MO
SON 64/G3 56.35 19.03 14.21 - 0.15 2.83 - - - 1.22 0.31 - 1.06
SON 68 18 17LIC2A2ZI 5.47 13.96 11.03 4.59 3.34 1.26 4.99 - 2.13 1.49 0.28 0.59 0.82
SAN 60 46.49 15.71 11.08 10.77 11.42 0.13 4.02 - - 0.08 0.02 - 0.07
SM 58 LW 11 58.70 10.94 8.31 7.76 0.71 0.45 4.24 3.15 3.45 - 0.57 0.26 0.13 0.51
ABS 29 57.47 15.16 10.07 6.67 1.63 0.25 - - - 4.90 0.69 0.20 0.59 0.75
ABS 41 58.31 15.39 10.22 6.76 1.65 1.27 2.48 0.70 0.20 0.60 0.76
ABS 39 57.09 19.40 14.72 - 2.15 2.52 _ _ _ - 0.73 0.21 0.63 0.80
ABS 118 52.45 13.93 11.06 4.64 3.33 1.26 4.94 - = 2.13 1.47 0.26 0.77 0.98
JSS-A 52.19 14.37 10.97 4.58 3.32 1.26 4.97 - 2.15 1.48 0.27 0.57 0.82
JW-D 54.86 18.14 12.39 4.25 1.33 1.51 2.27 - - 1.58 0.28 0.20 0.82
GC-12/9B 52.35 13.10 9.96 7.75 2.80 2.23 3.04 2.53 4.36 - 1.11 - 0.03 -
Average (when present) 52.55 12.77 11.82 6.87 3.19 2.56 3.22 3.20 2.04 1.82 0.97 0.75 0.83 0.64
Average (all glasses) 52.55 12.77 11.82 5.21 3.19 2.56 2.33 1.77 0.99 0.94 0.94 0.72 0.66 0.53
Median 52.47 12.94 11.08 7.01 2.55 1.51 3.04 2.34 1.80 2.13 0.75 0.21 0.59 0.78
Minimum 37.80 4.82 4.34 2.66 0.15 0.04 0.61 0.70 0.12 0.02 0.05 0.02 0.03 0.01
Maximum 58.70 21.01 25.31 10.77 11.42 6.13 6.22 10.02 4.63 4.90 2.71 8.20 3.02 1.33
Occurrence (100 100 100 76 100 100 72 55 48 52 97 97 79 83
I M M- - I- I-
Table 4-3. Compositions of selected candidate waste glasses (mol%) (Ellison et aL, 1994) (cont'd)
Glass KIONO CeO Nd20 3 P20 ICsO RuO. Sro Cr2O Th% L&203 U30. PdO PrTI3
HW-39-1 - 0.53 0.08 0.10 _ 0.05 0.05 0.06 0.57 - 0.10 - - 0.02
SRL 131A 2.88 1.17 - - - - 0.01 0.06 - 0.00 0.23 - -
SRL 165A 0.13 0.75 _ - - - 0.07 - - - 0.07 - _
SRL 202A 2.67 0.74 _ - - - - 0.02 0.04 0.07 _ 0.16 - _
SRL 211 0.04 1.53 0.35 0.19 - - 0.24 0.00 - - - -
WVCM-47 0.99 0.27 0.03 0.03 1.28 0.02 - 0.02 0.10 0.99 0.01 0.05 - 0.01
WVCM-50 1.25 0.30 0.31 0.03 1.28 0.02 - 0.01 0.07 0.98 0.01 0.05 - 0.01
WVCM-70, Ref-6 3.78 0.24 0.07 0.03 0.60 0.02 0.04 0.01 0.07 0.96 0.01 0.05 0.02 0.01
WV 205 2.55 0.64 0.06 - 1.21 0.02 - - 0.10 - - - - _
PNL 76-68 - 0.25 0.44 0.90 0.36 0.22 0.43 - 0.24 - 0.11 - - -
GP98/12 - 0.08 0.16 0.07 0.20 0.05 0.25 - 0.02 - 0.02 0.02 0.17 0.02
UK 209 - 0.31 0.37 0.34 0.10 0.17 0.33 0.20 0.23 - 0.09 - 0.23 0.08
UK 189 - 0.31 0.37 0.34 0.11 0.17 0.32 0.20 0.23 - 0.09 - 0.22 0.08
C 31/3 _ 0.25 0.62 0.58 _ 0.49 0.56 0.46 0.22 - 0.16 0.04 0.07 0.15
B 1/3 - 0.26 0.62 0.58 - 0.49 0.32 0.46 0.23 - 0.16 0.05 0.05 0.15
VG 98/3 - 0.20 0.64 0.45 0.30 0.34 0.56 0.34 0.11 - 0.13 0.10 0.36 0.12
SON 58 - 1.13 0.93 0.68 0.33 0.49 1.08 0.49 0.10 - 0.46 0.33 0.61 0.21
SON 64 - 0.23 0.54 0.40 - 0.29 034 0.29 0.20 0.12 0.13 0.04 0.20 0.12
Table 4-3. Compositions of selected candidate waste glasses (mol%) (Ellison et al., 1994) (cont'd)
FGlass 0 NiO CeO0 I Nd2O P0 CS20 RuO2 SrO I Cr.0 ThO2 LaO0 UO. PdO PO
SON 64/G3 - 0.21 0.58 0.42 0.22 0.31 0.37 0.10 0.25 - 0.14 0.08 0.21 0.13
SON 68 18 17 L1C2A2ZI - 0.69 0.37 0.33 0.14 0.32 - 0.23 0.23 0.09 0.19 0.04 - 0.09
SAN 60 - - 0.04 0.01 - - - 0.03 - - 0.04 - - 0.02
SM 58LW I1 - 0.09 0.23 0.39 - - - - - - - - - -
ABS 29 - 0.33 0.29 0.24 - 0.21 - 0.17 - - 0.14 0.13 - 0.08
ABS 41 - 0.33 0.29 0.24 - 0.21 - 0.17 - - 0.15 0.13 - 0.08
ABS 39 - 0.35 0.31 0.25 - 0.22 - 0.18 - - 0.15 0.14 - 0.08
ABS 118 - 0.80 0.38 0.31 0.15 0.27 - 0.22 0.23 - 0.19 0.07 - 0.10
JSS-A - 0.69 0.37 0.33 0.23 0.35 - 0.22 0.23 0.04 0.19 0.04 - 0.09
JW-D - 0.30 0.56 0.33 - 0.24 - 0.22 - - 0.11 - - 0.10
GC-12/9B 0.01 0.07 - 0.15 - - - 0.06 0.05 - - 0.38 - -
Average (when present) 1.59 0.47 0.36 0.31 0.46 0.23 0.39 0.18 0.16 0.46 0.12 0.11 0.21 0.08
Average (all glasses) 0.49 0.45 0.31 0.27 0.22 0.17 0.16 0.16 0.12 0.11 0.10 0.08 0.07 0.06
Median 1.25 0.31 0.37 0.33 0.27 0.22 0.33 0.17 0.16 0.12 0.13 0.07 0.21 0.08
Minimum 0.01 0.07 0.03 0.01 0.10 0.02 0.04 0.01 0.00 0.04 0.00 0.02 0.02 0.01
Maximum ~~~~~~3.78 1.53 0.93 0.90 1.28 0.49 1.08 0.49 0.57 0.99 0.46 0.38 0.61 0.21
Occurrence (%) ~3 1 97 86 86 48 76 4 1 90 76 24 79 69 3 72
I.I.-.110
I M 0-
Table 4-3. Compositions of selected candidate waste glasses (mol%) (Ellison et al., 1994) (cont'd)
Glass Gd2O3 TeO2 Y203 Te2O3 J CuO Sm 2O3 J CoO AsI3Ag 2 O
HW-39-1 _ _ 0.01 - 0.08 0.01 - _
SRL 131A _ - - 0.01 0.02 - _
SRL 165A - - - -
SRL 202A - - - 0.01 0.34 _ - _ _
SRL211 - - - _ 0.00 - _ _ _
WVCM-47 - 0.01 - - 0.01
WVCM-50 - - 0.01 - 0.04 - -
WVCM-70, Ref-6 - - 0.01 - 0.03 0.01 0.02 _
WV205 - - - -
PNL 76-68 0.01 - - - - _ _
GP 98/12 - 0.04 0.02 - - 0.01
UK209 - - 0.05 - - - _
UK 189 - - 0.05 - - -
C31/3 - 0.13 0.12 - - - _ 0.20 _
B 1/3 - 0.13 0.12 - - - 0.16 _
VG 98/3 0.01 0.11 0.10 - - - _ _ _
SON 58 0.02 0.24 0.13 0.21 - - _ 0.01
SON 64 0.01 0.14 0.07 0.12 - - 0.01
0
Table 4-3. Compositions of selected candidate waste glasses (mol%) (Ellison et al, 1994) (cont'd)
Glass GdO2L TeO 2 Y203 TcO2 . CuO Sm2r 3 E CoO As2O3 Ag2
SON 64/G3 1.25 0.15 0.08 0.13 - 0.01
SON68 18 17LIC2A2ZI - 0.10 0.06 - 0.11 0.01
SAN 60 - - - 0.04 - -
SM 58 LWI - - - - - - 0.09
ABS 29 - - 0.04 - - - -
ABS 39 - - 0.05 - - -
ABS 118 - - 0.06 - - - - -
JSS-A - 0.10 0.06 - - - 0.11 0.01
JW-D 0.01 0.10 0.06 - - 0.06 0.11 0.10
GC12/9B - - - - - - - - -
Average (when present) 0.22 0.12 0.06 0.09 0.08 0.05 0.09 0.18 0.02
Average 0.05 0.04 0.04 0.02 0.02 0.02 0.02 0.01 0.00
Median 0.01 0.12 0.05 0.08 0.03 0.06 0.11 0.18 0.01
Minimum 0.01 0.04 0.01 0.01 0.00 0.01 0.02 0.16 0.01
Maximum 1.25 0.24 0.13 0.21 0.34 0.13 0.11 0.20 0.10
Occurrence (%/) 21 34 69 21 21 31 17 7 21
.P-
Figure 4-2. Schematics of the silica glass network (Varshneya, 1994)
4.2.3 Alkali-Alkaline Earth-Silicate Glasses
Alkaline earth cations (N) are bivalent, similar to alkali ions, behave like network modifiers, andoccupy interstitial sites in the glass network. The addition of one alkaline earth ion creates two NBO sitesas shown by Eq. (4-2).
_Si- -Si +NO Si- (4-2)
Compared to alkalis, which have a significant effect on glass properties, the alkaline earth ions tendto provide stability to the glass structure.
4.2.4 Alkali Borate Glasses
B203 is a glass former oxide, and its basic structural unit is the B0 3 triangle. The addition of an alkalioxide could result in (a) creation of an NBO as shown by Eq. ( 4-3) or (b) conversion of a 3-coordinatedboron (B3) state to a 4-coordinated boron (B4) state as shown by Eq. (4-4).
-B-O-B=+M 2 0 BB-O0-M+M+O-B- (43)
=B-O-B=+M 2 0 [-B-O0]-M+M+-O-B-] (44)
4-12
As alkali ions are added, B3 states are converted to B4 states. The extra negative charge is satisfiedbythe alkali ion while increasing the connectivity ofthe system (increase in coordination number). Increasedconnectivity results in an increase in the viscosity and decrease in thermal expansion. Since the negativecharge is distributed over a large [BO4 group (not localized as in the case of NBO where an alkali ion isattached only to oxygen), the alkali ion is expected to be more mobile than when in the B3 state.
Figure 4-3 shows the effect of adding Na2O to B203. Curve "c" shows the thermal expansionbehavior, which indicates a minimum at 16 wt% Na2Q. Biscoe and Warren (Varshneya, 1994) suggestedthat the addition of each alkali ion (up to 16 wt%) to boron oxide causes one boron to change from the B3
state to the B4 state with no creation of NBOs causing thermal expansion to decrease. An alkali oxidegreater than 16 wt% causes production of NBOs, which increases thermal expansion. In 1963, Bray andO'Keefe (Varshneya, 1994) experimentally determined the fraction N4 of B4 in alkali borates. Figure 4-4shows their results. For alkali borate composition xNa2O(l -x)B2 03, a fraction of N4 is given byx/(l -x).This behavior is shown by the solid line in figure 4-4. N4 increases with the increase in alkali content but,instead of stopping at 16 mol% alkali, it continues up to 45 mol% and decreases thereafter. The N4
concentration does drop go to zero until 70 molalkcali. Brayand O'Keefe confirmed Biscoe and Warren'sargument of change in coordination number; however, the B anomaly occurring at 16 wt% alkali oxidecannot be explained merely by the change in coordination number. (This behavior is known as a boronanomaly). Kuppinger and Shelby (1985) showed an anomaly to occur over a much wider range (up to
4_-
co
(D0
6
0
4 x
,o
.012 =000
0cLcox rco
D 20 30Percentage Na2O
Figure 4-3. Thermal expansion coefficient and density of Na20-B2 03 glasses. Curve a: calculateddensity, curve b: measured density, and curve c: measured thermal expansion coefficient(Varshneya, 1994)
4-13
0.5
0.4 0_
0.3 0 4A A
z
0.2 es-x
0.1 /4+ A
0 10 20 30 40 50 60 70
Alkali (Mol %)
Figure 4-4. The fraction N4 of 4-coordinated boron atoms in alkali borate glasses as a function ofmol% alkali. Na 2O (D); K20 (0); Li20 (A); Rb2O (+); Cs20 (X). (Varshneya, 1994)
30 wt%). Currently, aboron anomalyis regarded as amanifestation ofthe various boron oxide groups, suchas boroxol, pentaborate, triborate, or diborate, present in the system.
The number of NBOs in the alkali borate system can be calculated using Gupta's rule (Varshneya,1994). If the alkali borate compositionis expressed as 7Na2O B 203 thenN 4 = T, andfNB0 = 0 upto T= 0.5.When T > 0.5, then N4 is given by (3 - 7)/5 andfNBO is calculated by alkali ions not used by N4.
4.2.5 Alkali Borosilicate Glasses
Alkali borosilicate glasses have two network formers-B and Si-and glasses are represented bya general formula of TNa20KSiO 2.B203 . The addition of an alkali ion either creates an NBO with Si orB. Nuclear magnetic resonance studies have indicated that alkali ions tend to associate with the boron aslong as T< 0.5. Beyond T= 0.5, alkali distribution is partitioned depending onthe ratio of K, whichis definedas [SiOJ/[B2 03 1 and shown in figure 4-5.
4-14
1.i
0.8 -
0.6 -
0.4 0.8 1.2 1.6 2.0 2.4
R = [Na2O] / [B203 ]
Figure 4-5. Fraction of tetrahedrally coordinated borons (N4) as a function of mol Na2O/B203 andSiO2 ratios determined using nuclear magnetic resonance and molecular dynamics simulation(Varshneya, 1994)
4.2.6 Alkali Aluminosilicate Glasses
The structural configuration of Al3+in alkali aluminosilicate glass depends onthe ratio ofAl2 03 /M2 0.
Whenthe A1203/M20 ratio is <1, the Al3ion goes inthe glass as atetrahedrallycoordinated network formersimilar to the boron B4 coordination discussed earlier. The excess negative charge on the [A10 41- group issatisfied bythe alkaliioninthe system. Thus, the addition ofoneAI3 +ionto alkali silicate glass removes oneNBO. For A120 3/M 2 0 > 1, the A3+ ion goes inthe network as amodifier in an octahedral coordination. Thenumber ofNBOs are calculated by associating each alkali ion with A13+ ion. If the number of alkali ions isgreater than Al3+ ions, the alkali ions create NBOs with Si. If the A13+ ions are greater than alkali ions, alkaliions are associated with A13+ and the remaining A13' ions create three NBOs.
4.2.7 Alkali Aluminoborosilicate Glasses
In alkali aluminoborosilicate glasses, there are three glass formers: Al, B, and Si. Darab et al. (1996)showed, using magic-angle spinning nuclear-magnetic resonance, that for A1203/M 20 <1, the preferredsequence of alkali ions attaching to various glass formers is Al, B, and Si.
4-15
4.2.8 Fe3" in Alkali Silicate, Aluminosilicate, and Borosilicate Glasses
Fe3" ions, similar to AlP ions, exist in four-fold coordination and behave like a glass former. But underreducing conditions, Fe3 ' reduces to Fe2" ion and enters the structure as a network modifier. Figure 4-6shows the dependence of Fe2+/Fe(tot) on K20/(A1203+K20) in alkali alumino silicate glasses (Dickensonand Hess, 1986). For K20/(A1203+K 20) < 0.5, there are not enough K+ ions to charge balance all AI3+ ions.Therefore, Fe as Fe2+ ions compete for charge balancing. Therefore, Fe2+/Fe(tot) increases sharply with thedecrease in K20/(AI203+K2O) ratio. For K20/(A120 3+K 20) > 0.5, there are sufficient K+ ions to chargebalance all Al3l ions and some Fe2+ ions depending on the K+ ion concentration. This study indicates that,in alkali aluminosilicate glasses with Fe, the preferred sequence of alkali ions attaching to various glassformers is Al, Fe, and Si.
The role of Fe3+ ions in sodium borosilicate glasses containing up to 12 wt% Fe2O3 has been studiedbyMagini etal. (1984), Licheri et al. (1985) and Agostinelli et al. (1987). The studies showedthepresenceof Fe3+ ions in four-fold coordination suggesting that Na' ions balance the charge with Fe3` ions before B3+ions.
0.50
U1)U-
U)
0.40 0.50 0.60 0.70 0.80
K 2 0/(K2 0 + A120 3 )
Figure 4-6. The ratio of Fe2+/Feto, in (20-x) K20xAI203 8OSiO2 containing approximately 1 mol%Fe2O3 (Dickenson and Hess, 1986)
4-16
DeYoreo et al. (1990) determined stabilizationenthalpies forvarioustrivalent cations and showed thatA13+ ions are most effective followed by Fe31 ions and B3+ ions in the structure for charge balancing.
4.3 USE OF NONBRIDGING OXYGEN IN DEVELOPING NUCLEAR WASTE
GLASS PROPERTY MODELS
Glass properties, such as viscosity, electrical conductivity, thermal expansion, Tg, and TL, areinfluenced bythe connectivity ofthe glass structure, which is determined bythe number ofNBOs. The lowerthe number of NBOs, the higher the glass structure connectivity. Determining the number of NBOs in amulticomponent system having four or five glass-forming oxides requires experimental studies to determinepreferred sequencing or partitioning of alkali and alkaline earth oxides among various glass-forming oxides.Jantzen (1991) developed electrical conductivity, viscosity, and liquidus models based on the number ofNBOs in the glass structure. In a multiple-component system, each component could either add or reduceNBO. Jantzen (1991) assumed the following rules to determine the total NBOs in a glass melt:
* A1203 forms A104 tetrahedral linkages and two BO bonds for each A1203 present.
* Fe203 can form BO and NBO bonds depending on the amount of M2 0 and A1203.
* For glasses with A1203/M20 < 3/4, low concentrations of B203 enter the glass structure as(BOQ tetrahedra, while at higher B203 concentrations these tetrahedra are converted intoplaner (BO3)- groups containing one NBO.
The total number of NBO in a given melt composition, is calculated by Eq. (4-5)
NBO = 2(M2 0 + Fe 203 - A1203) + B203 (45)SiQ 2
where M2 0 = Na2O + K20 + Li20 + Cs 20
Based onrecent studies noted insection4.2.8, the association of Fe2 03 inthe glass structure dependson various other components. Therefore, the assumption that each mol of Fe203 creates 2NBO may onlybe partially valid. It is advised to evaluate the number of NBOs to glass composition using rules discussedbelow.
CombiningNBO formation rules provided inthe standard glass sciencetextbook(Varshneya, 1994),and in papers previouslydiscussed,thefollowingrules canbeused forestimatingNBOsinmulticomponentglasses:
* All calculations are made on a molar basis.
* Each alkali ion creates one NBO.
* Each alkaline-earth ion creates two NBOs.
4-17
* A boron ion can exist either in a 3-coordination state (B3) or 4-coordination state (B4). For eachalkali (Mt) ion added, one B from a B3 state converts to a B4 state without creating any NBOprovided T, which is defined as the ratio of M2O/B203 < 0.5. If T> 0.5, the number of B4 statescan be estimated using Gupta's approximation N4= (3 - 7)/5 (Varshneya, 1994).
* In alkali borosilicate glasses, there are two glass formers: Si and B. The alkali could eitherassociate with Si creating an NBO or with B converting a B3 state to a B4 state and creatingno NBOs. Studies indicate that alkalis prefer to associate with B as long as T < 0.5. Theremaining alkali are attached to silicon.
* In alkali aluminosilicate glasses, A120 3/M 20 < 1, Al3+ ions go in as a network former andremove one NBO. If A120 3/M 2 0 > 1, A13' enters the network as a modifier in octahedralcoordination with three oxygen as NBO and three BO. Remaining alkali ions are attached toSi.
* In alkali aluminoborosilicate glasses, there are three glass formers: Al, B, and Si. Darab et al.(1996) showed that for A120 3/M 20 < 1, the preferred sequence of alkali ions attaching tovarious glass formers is Al, B, and Si.
* In alkali aluminoborosilicate glasses containing iron, the preferred sequence of alkali ionsattaching to various glass formers is Al, Fe, B, and Si.
* The coexistence of Fe as Fe2 ' and Fe3' requires that Fe2+ should be considered along with analkaline earth ion, and Fe31 should be considered along with trivalent ions.
4.4 SUMMARY
This chapter provided a brief review of glass structure and glass forming principles and systems.While the discussion was focused on simple systems, complex systems such as nuclear waste glasses followsimilar glass formation rules. The information provided in this chapter is used in the following chapters toexplain various properties of the nuclear waste glasses.
4.5 REFERENCES
Agostinelli, E., D. Fiorani, and E. Paparazzo. XPS Studies of iron sodium borosilicate glasses. Journal ofNon-Crystalline Solids 95 and 96: 373-380. 1987.
Darab, J. G., X. Feng, J. C. Linehan, and P.A. Smith. Composition-structure relationships in model Hanfordlow-level waste glasses. Proceedings of the Environmental Issues and Waste ManagementTechnologies for the Ceramic and Nuclear Industries II Symposium. 9 8 th Annual Meeting ofthe American Ceramic Society, Indianapolis, Indiana, April 14-1 7, 1996. Ceramic TransactionsVolume 72. V. Jain and D. Peeler, eds. Westerville, OH: American Ceramic Society: 103-110.1996.
4-18
DeYoreo, J.J., A. Navrotsky, and D.B. Dingwell. Energetics of the charge-coupled substitution Si4+ ->Na+ + T3+ inthe glasses NaTO2-SiO2 (T=/Al,Fe,GaB). Journal of the American Ceramic Society73(7): 2,068-2,072. 1990.
Dickenson, M.P., and P.C. Hess. The structural role and homogeneous redox equilibria of iron inperaluminous, metaluminous and peralkaline silicate melts. Contrib. MineralPetrol. 92: 207-217.1986.
Ellison, A.J., J.J. Mazar, and W.L. Ebert. Effect of Glass Composition on Waste Form Durability: ACritical Review. ANL-94/28. Argonne, IL: Argonne National Laboratory. 1994.
Jantzen, C.M. First principles process-product models for vitrification of nuclear waste: Relationship of glasscomposition to glass viscosity, resistivity, liquidus temperature, and durability. Proceedings of theFifth International Symposium on Nuclear Waste Management IV 93 rd Annual Meeting of theAmerican Ceramic Society, Cincinnati, Ohio, April 29-May 3, 1991. Ceramic TransactionsVolume 23. G.C. Wicks, D.F. Bickford, and L.R. Bunnell, eds. Westerville, OH: American CeramicSociety: 37-51. 1991.
Kuppinger, C.M., and J.E. Shelby Viscosityand thermal expansion of mixed-alkali sodium-potassium borateglasses. Journal of American Ceramic Society 68(9): 463-467. 1985.
Licheri, G., G. Paschina, G. Piccaluga, G. Pinna, M. Magini, and G. Cocco. On the coordination of iron ionsin sodium borosilicate glasses. EXAFS Investigation. Journal of Non-Crystalline Solids 72:211-220. 1985.
Magini, M., A.F. Sedda, G. Licheri, G. Paschina, G. Piccaluga, G. Pinna, and G. Cocco. On the coordinationof iron ions in sodium borosilicate glasses. Wide Angle X-Ray Diffraction Investigation.Journal of Non-Crystalline Solids 65: 211-220. 1984.
Paul, A. Chemistry of Glasses. New York: Chapman & Hall. 1982.
Rawson, H. Inorganic Glass-Forming Systems. New York: Academic Press, Inc. 1967.
Varshneya, A.K Fundamentals of Inorganic Glasses. New York: Academic Press, Inc. 1994.
4-19
5 ELECTRICAL CONDUCTIVITY
The electrical conductivity in most oxide glasses is attributed to migration of positively charged alkali ions inthe presence of an applied electric field. Because alkali ions are weakly linked with the glass structure, theyare relatively more mobile than other ions and become carriers ofthe electrical current. Electrical conductivityis influenced by the size of the ions and the strength of their bonds in the glass structure. An understandingofthe electrical conductivity of glass melts and its temperature and compositional dependence is importantfor the operation of joule-heated melters. Glass compositions are designed to provide specific electricalconductivity at melting temperatures. In addition, electrical conductivity data are used to estimate powerrequirements for a joule-heated melter. Even though electrical conductivity is a process parameter that isdefined bythe glass composition, processing conditions could lower electrical conductivitybythe formationof metallic orhighlyconductive species inthemelt. Inthis chapter, abrief discussion on electrical conductivitylimits imposed onthe glass compositionis followed byareview of various compositional studies on electricalconductivity and models correlating composition to electrical conductivity and the impact of melter conditionson electrical conductivity.
5.1 ELECTRICAL CONDUCTIVITY LIMITS
The joule-heated melters for nuclear waste vitrification require electrical conductivity of the glassmelts to be between 10 and 100 S/in at the melting temperature (Hrma et al., 1994).
The lower limit of 10 S/m. for electrical conductivity is imposed on the glass melts at meltingtemperature to ensure the electrical conductivity of the glass melt is significantly higher than the electricalconductivity of the refractories surrounding the glass melt. The electrical conductivity of the high-chromerefractory at 1,150 'C is 0.5 S/m, which is a factor of 20 lower than the glass melt electrical conductivity. Ifthe electrical conductivity of the melt is below 10 S/m, it could require higher voltage to maintain desiredoperating conditions in the melter. This higher voltage could cause the current to flow through the refractory.
The upper limit of 100 S/in for electrical conductivity is imposed on the glass melt at meltingtemperature to ensure the glass melt provides enoughjoule heating for melting. Highly conductive glass meltswill require high current density for melting and could exceed the maximum operating current density of3 x 104 A/m2 specified for Alloy 690 electrodes (Wang et al., 1996).
For thejoule heating of glass, alternating current is used because the electrolytic properties of glassare such that the use of direct current causes polarization in the melt with alkali ions moving to cathode andoxygen being liberated at the anode. In addition, direct current causes formation of deposits and blisters onthe electrode surfaces.
5.2 ELECTRICAL CONDUCTIVITY IN GLASS MELTS
The electrical conductivity in most ofthe borosilicate-based waste glasses is attributed to positivelycharged alkali ions. Because alkali ions are weakly linked with the glass structure, they are relatively moremobile than other ions and become carriers ofthe electrical current. The electrical conductivity is influencedbythe size ofthe ions and the strength of their bonds in the glass structure. Between Na and K ions, K+ ionsare bound less strongly, but because oftheir larger size face ahigher resistance during migration. The Li ions,
5-1
on the other hand, are smaller in size than Nae ions but are more strongly bound to the glass structure, andface more resistance than Nae ions. Therefore, Nae ions are the principal contributor to electrical conductivityand are the most mobile ions in the glass melts. Glasses containing a mixture of alkali ions exhibit a nonlinearbehavior in which the conductivity passes through a pronounced minimum as one alkali ion is substituted forthe other in the glass composition. This behavior is referred to as the mixed-alkali effect. Figure 5-1 showsthe mixed-alkali effect in a mixed Na2 O and K20 aluminosilicate glass melt (Kim, 1996). The minimum inelectrical conductivity becomes less pronounced at higher temperatures.
The electrical conductivity increases as temperature increases. The temperature dependence oftheelectrical conductivity of a glass melt is Arrhenius in nature and can be expressed as
E.
E = 'Oe RT(5-1)
where, so is a constant, E, is activation energy, R is a gas constant, and T is the temperature in K.Equation (5-2) can be rewritten as
log(s) = a + bT
(5-2)
c-)E
-C
0)0
K 2 0/M 2 0
Figure 5-1. Conductivity (a) isotherms for 25R2 0-10A1203 mixed-alkali melts against K2O/M2Oat five temperatures (0) 1,000 'C; (+) 1,100 'C; (-) 1,200 'C; (-) 1,300 'C; and (E) 1,400 'C(M 20 = Na2O + K2O mol%) (Kim, 1996)
5-2
where
a = logs, (5-2a)
and
b= E (5-2b)2.303R
Electrical conductivity of glass melts is measured using an experimental configuration as shown infigure 5-2 (Eaton et al., 1991). The apparatus consists of a furnace, in which a platinum or alumina cruciblefilled with atest sample is placed, and aprobe head, which consist oftwo electrodes and a thermocouple. Theprobehead is lowered into a glass melt to a predetermined height. The electrodes are either rods or platesmade of Pt-Rh alloy. Electrodes are connected to an impedance analyzer. Electrical conductivity is measuredby applying current frequency ranging from 0.1 to 100 kHz and measuring electrical resistance between theelectrodes. In glass melts, no frequency dependence is observed at frequencies >1 kHz. Since the cellgeometry is not specifically defined, a cell constant is obtained by measuring electrical conductivity ofstandard KCI solution and verified by measuring electrical conductivity of a standard glass. Electricalconductivity is determined by Eq. (5-3).
L/A (53)R
where LIA is a cell constant, and R is a measured resistance at a given frequency.
Thermocouple
Pt-Rod Electrodes
Crucible r
Figure 5-2. Schematic of an electrical resistivity apparatus (Eaton et al., 1991)
5-3
5.3 ELECTRICAL CONDUCTIVITY DATA AND MODELS
Hrma et al. (1994) conducted a multiyear, statistically designed compositional variability study (CVS)to characterize the relationship between composition and properties. Table 5-1 shows target composition, andthe upper and lower limits for various components in the study. The target composition, HW-39-4, is basedon the proposed Hanford HLW glass composition for the neutralized current acid waste (NCAW) (nowknown as Envelope B/D waste) (Calloway et al., 1999). Electrical conductivity was measured on 120 glassesat three to five different temperatures between 950 °C and 1,250 'C. The electrical conductivity at 1,150 'Cwas determined by Arrhenius fit.
Compositional dependence on electrical conductivity in S/m, at 1,150 'C was determined by thefirst-order mixture models as shown by Eq. (5-4).
10
ln61150 E bjxiI
(5-4)
where Es150 is electrical conductivity at 1,150 'C, xi and b, are mass fraction and regression coefficients forcomponent, respectively, i. The estimated regression coefficients are shown in table 5-2. Hrma et al. (1994)also used second-order mixture model with first-order mixture model terms and several second-orderterms,which were selected by using statistical variable selection techniques to improve the relationship betweenelectrical conductivity and temperature and composition. The second-order mixture model did not fit theelectrical conductivity data substantially better than the first-order mixture model.
Based on the regression coefficients provided in the table 5-2, a component effects plot wasdeveloped centered around the composition HW-39-4 (figure 5-3). The analysis predicts Li20 and Na.0have the strongest effects in increasing the electrical conductivity while SiO2 has the strongest effect indecreasing the electrical conductivity.
Table 5-1. Composition region (mass fraction) studied by Hrma et al. (1994)
HW-39-4 Lower Limit Upper LimitSiO2 0.5353 0.42 0.57
B203 0.1053 0.05 0.20
Na2O 0.1125 0.05 0.20
Li2O 0.0375 0.01 0.07
CaO 0.0083 0.00 0.10
MgO 0.0084 0.00 0.08
Fe203 0.0719 0.02 0.15
Al203 0.0231 0.00 0.15
ZrO2 0.0385 0.00 0.13
Others 0.0592 0.01 0.10
5-4
Table 5-2. Regression coefficients for In v1150 (Hrma et al., 1994)
I Xi | bi
SiO2 0.8471
B20 3 2.2581
Na2 O 11.0396
Li2O 23.5355
CaO 1.4129
MgO 1.0565
Fe2O3 2.5863
A120 3 1.3108
ZrO2 1.1224
Others 3.453 1
Fu (1998) analyzed electrical conductivity data on 43 simulated and actual waste glasses from sixdifferent projects. The selected glasses were required to (i) contain a minimum of five cations from severalprojects, (ii) contain <1 mol% anions other than 02-, (iii) have the analyzed and calculated composition toagree within 10 relative percent, and (iv) contain only one phase. The conductivity of these glasses rangedfrom 0.6 to 130 S/cm. The composition range is shown in table 5 -3. In figure 5 -4, electrical conductivity isplotted as a function oftotal mols of alkali oxide disregarding the concentration ofindividual alkali oxide. Thefollowing empirical equation was obtained using a quadratic fit with r2 of 0.94.
Ins 1150 = 0.282 x (M 20) - 0.00373 x (M 2 0)2 - 5.34 (5-5)
where M20 is mol% total alkali oxide, and s is electrical conductivity in S/cm. For 25 new compositions, thismodel predicted electrical conductivity within 25 relative percent error, but the model failed to predictelectrical conductivity in glasses containing a sodium sulfate phase. In contrast to the conclusion reached byHrma et al. (1994) that Li2O affects the electrical conductivity more stronglythanNa2O, this model assumesthat all alkali contribute equally and that the mixed alkali effect plays a minor role in the determination of meltconductivity for multicomponent oxide glasses.
Jantzen (1991) developed an electrical-resistivity model based on the number ofNBO in the glassstructure. The total number of NBO is a given melt composition, is calculated by Eq. (4-5). Based on theNBO calculations, an empirical relationship, shown in Eq. (5-6), was developed. The relationship providedan YF = 0.57 indicating significant scatter in the data.
logpI150 = -0.48NB0 + 0.648 (5-6)
5-5
EC)
0
0LO
Ii>%
C.)
-o-0C00
ci,8
w-oii]
ai)a.
60 1-
50
401
, Na 2O
/20
I Isio N / /
N.N /N\ / ,/
NIIN-' /7/
B 0 C aO ~2 ZrO /
Fe2O3. . . . . OthersOthers MO -'l. - eO
A1203
Na2 0 /O
30 1
20 I
-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16Component Wt% Change from HW-39-4 Value
Figure 5-3. Predicted component effects on electrical conductivity at 1,150 'C relative to theHW-39-4 composition, based on the first-order mixture model using mass fractions (Hrma et al.,1994)
5-6
Table 5-3. Compositional ranges of six multicomponent systems and their viscosity andconductivity ranges at 1,150 °C (Fu, 1998)
Composition (mol%)
Components Series I Series II Series III Series IV Series V Series VI!
A1203 7.0 5.0-6.8 6.2-10.6 0-10.7 4.3-6.8 7.0-12.4
B203 9.7 9.6-14.9 9-15/2 9.58 6.1-9.6 11.2-12.7
BaO 0.0 0.0 1.12-1.3 0.0 0.0 0.01
CaO 35-0 9.2-12.1 8.1-17.3 25.35 16.2-25.4 0.2-0.4
Fe 203 0.0 0.0 1.7-4.6 0.0 0-10.1 1.4-5.8
K20 0.0 0.0 0.5-0.9 0.0 0.0 0.6-1.1
Li2O 0.0 0.0 0.4-4.5 0.0 0.0 0.1-7.7
MgO 35-0 0.0 4.8-8.9 0.0 0.0 0.2-6.4
Na2O 0.0 19.6-23.0 7.1-8.1 17.45-6.75 10.7-32.8 15-28
NiO 0.0 0.0 0.1-0.4 0.0 0.0 0.7-2.0
P205 0.0 0.0 1.1-3.3 0.0 0.0 1.2-2.1
PbO 0.0 0.0 0.7-1.8 0.0 0.0 NA
SO3 0.0 0.0 0-1.3 0.0 0.0 NA
SiO2 48.9 46.3-48.4 39.3-43.5 47.62 30.4-47.6 39.9-51.1
U3 08 0.0 0.0 0.0 0.0 0.0 0.2-0.5
ZrO2 0.0 1.6-2.1 0-0.3 0.0 0.0 0.01-3.7
Total 100.0 100.0 100.0 100.0 100.0 100.0
Viscosity range 30-72 30-73 53-300 15-135 9-58 45-1,550(poise) at1,150 C
Conductivity 0.006-1.3 0.24-0.76 0.02-0.08 0.08-1.09 0.08-1.1 0.17-0.62range (S/cm) at1,150 °C
NA = Not analyzed
5-7
1
o n c|
0 0~~~~~~V ~V V 0
-1 ° O U
: -2
O Xo + Xo -3_ ° + + +xx
4 X-4 0~~~~~~~~~ SerieslI V Series 11 x Series III
-5 + Series IV o Series V Series VI
-6 I I I0 5 10 15 20 25 30 35
Alkali (Mol %)
Figure 5-4. Conductivity as a function of alkali content at 1,150 'C (Fu, 1998)
where p is defined as electrical resistivity in Q-cm. Combination of compositional dependence andtemperature yielded the relationship shown in Eq. (5-7) with r2 = 0.92.
logp = -2.48 + 4 -9 0.45NBO (5-7)TQ'K)
DWPF uses Eqs. (5-6) and (5-7) to estimate electricalresistivity(reciprocal ofelectrical conductivity)of glass melts. Jantzen's model, similar to Fu's model, assumes all alkalis contribute equally to electricalconductivity. Based on Hrma' s experimental work, this assumption is not supported. The models reviewedin this section were developed for a specific range of glass compositions and glass components. Theapplicability ofthese models to other wastes, however, is limited. The user, however, is cautioned to performa careful and rigorous evaluation priorto using anymodel for predicting electrical conductivity ofglass melts.
5.4 EFFECT OF MELTER CONDITIONS ON ELECTRICAL CONDUCTIVITY
Even though electrical conductivity is a process parameter that is defined bythe glass composition,the processing conditions in the melter could lead to the lowering of electrical conductivity through theformation of a species causing electronic conduction in the melt. A species causing electronic conductioncould form eitherthroughthe accumulation ofnoble metals or the formation of highly conductive species suchas NiS, CuS, or FeS. The formation of such species results in electronic conduction in the melt, which isseveral orders of magnitude higher than ionic conduction; thus leading to electrical shorting in the melter.
5-8
5.4.1 Noble Metal Accumulation
Noble metals are present in the HLW as a result of radioactive decay, and they have a very limitedsolubility in the glass melt. Depending on the residence time of glass melt in the melter, noble metals eitherflush out with the molten glass during pour or settle on the melter floor. Since the failure ofthe Pamela Melterin Mol, Belgium, melters are designed to accommodate the accumulation ofnoble metals or are provided withdrains to flush noble metals. A detailed discussion on the subj ect is provided in Jain (2000). Monitoring thepower supply in the melter is a keyto monitoring the accumulation ofnoble metals inthe melter. An increasein the current between electrodes could signal an accumulation of conductive species in the melter.
5.4.2 Formation of Conductive Species
Extremely reducing conditions (Fe2 /Fe3+ > 1) in the melter could lead to the formation of highlyconductive species in the melter. In the presence of sulfur, conductive species such as FeS, CuS, and NiScould form at a much lower Fe2"/Fe 3" ratio. Since the waste compositions contain a significant amount oftransition elements, an effective redox control strategy, as discussed in chapter 9 of this report, is requiredfor maintaining redox within a specified range. An increase in current between electrodes could signalformation of conductive species in the melter.
Eventhoughthe electrical conductivityofthe targetglass composition does notpose anysafetyissue,the processing conditions inthe melter could lead to alowering of electrical conductivitythat could resultina short-circuit in the melter. A melter designed to accommodate noble metals, an effective redox controlstrategy, and administrative controls on monitoring the power supplyto the melter could reduce the risk andconsequences of a melter failure caused by a short circuit.
5.5 TANK WASTE REMEDIATION SYSTEM CONCERNS
The LAW glass composition contains a significantlyhigher concentration ofNa ions than HLW glasscomposition based on the proposed process flowsheet. The LAW glass composition could have higherelectrical conductivitythanHLWglass composition, but as long as both compositions fall within the prescribedelectrical conductivity range, there is little risk of processing problems in the melter. The risk caused by theformation of a conductive species or noble metal accumulation in the melter for LAW glass composition ismuch less glass composition compared to HLW. Compared to HLW, LAW composition contains a smalleramount oftransitionmetal elements that could form species that can cause electronic conduction inthe melterand have not noble metals.
5.6 SUMMARY
The electrical conductivity inmost oxide glasses is attributed to migration of positivelycharged alkaliions in the presence of an applied electrical field. Because alkali ions are weakly linked with the glassstructure, they are relatively more mobile than other ions and become carriers of the electrical current. Theelectrical conductivity is influenced by the size of the ions and the strength of their bonds in the glassstructure.
5-9
The joule-heated melters for nuclear waste vitrification require electrical conductivity of the glassmelts between 10 and 100 S/m at melting temperature (Hrma et al., 1994). The lower limit of 10 S/m forelectrical conductivity is imposed on the glass melts at melting temperature to ensure that the electricalconductivity of the glass melt is significantly higher than the electrical conductivity of the refractoriessurrounding the glass melt, which ensures the current does not flow through refractories. The upper limit of100 S/m for electrical conductivityis imposed onthe glass melt atmeltingtemperature to ensure the glass meltprovides enough joule heating for melting without exceeding the maximum operating current density forInconel 690 electrodes.
The electrical conductivity increases as temperature increases and is Arrhenius in nature. Thefirst-order mixture model, based on statistically designed experiments (Hrma et al., 1994), predicts Li2O andNa2 O have the strongest effects on increasing the electrical conductivity, while SiO2 has the strongest effecton decreasing the electrical conductivity. In contrast to the conclusion reached by Hrma et al. (1994) thatLi2O affects the electrical conductivity more strongly than Na2O, Jantzen (1991) and Fu (1998) concludedthat all alkalis contribute equally and that the mixed alkali effect plays a minor role in determining meltconductivity for multicomponent oxide glasses. The models reviewed in this section were developed for aspecific range of glass compositions and glass components. The applicability ofthese models to other wastes,however, is limited.
Even though electrical conductivity is a process parameter defined by the glass composition, theprocessing conditions in the melter could lower electrical conductivitybyforming metallic species inthe melt.The formation ofspecies causing electronic conduction could occur either through the accumulation of noblemetals or formation of highly conductive species, such as NiS, CuS, or FeS. The formation of such speciesresults in metallic conduction in the melt, which is several orders of magnitude higher than ionic conduction;thus leading to electrical shorting in the melter. A melter designed to accommodate noble metals, an effectiveredox control strategy, with administrative controls for monitoring the power supplyto the melter, could reducethe risks and consequences of a melter failure caused by a short-circuit.
5.7 REFERENCES
Calloway, T.B., Jr., C.A. Nash, D.J. McCabe, C.T. Randall, S.T. Wach, D.P. Lambert, C.L. Crawford,S.F. Peterson, and R.E. Eibling. Research and development activities in support of Hanfordprivatization-SRTC program (U). Proceedings of the Waste Management '99 Conference,Tucson, Arizona, February 28-March 4, 1999. CD-ROM Version. Tucson, AZ: WMSymposia, Inc. 1999.
Eaton, W.C., I. Joseph, andL.D. Pye. Somehigh-temperatureproperties of simulatedWestValleynuclearwaste glass. Proceedings of the Fifth International Symposium on Nuclear WasteManagement IV 93'd Annual Meeting of the American Ceramic Society, Cincinnati, Ohio, April29-May 3, 1991. CeramicTransactions Volume23. G.C. Wicks, D.F. Bickford, and L.R. Bunnell,eds. Westerville, OH: American Ceramic Society: 509-518. 1991.
Fu, S. Predicting melt conductivities of waste glasses. Proceedings of the International Symposium onEnvironmental Issues and Waste Management Technologies in the Ceramic and NuclearIndustries III, Cincinnati, Ohio, May 4-7, 1998. Ceramic Transactions Volume 87. D.K. Peelerand J.C. Marra, eds. Westerville, OH: American Ceramic Society: 253-260. 1998.
5-10
Hrma, P.R., G.F. Piepel, M.J. Schweiger, D.E. Smith, D.-S. Kim, P.E. Redgate, J.D. Vienna, C.A. LoPresti,D.B. Simpson, D.K Peeler, and M.H. Langowski. Property/Composition Relationships forHanford High-Level Waste Glasses Melting at 1,150 'C. Volume I Chapters 1-1 1. PNL-10359.Richland, WA: Pacific Northwest National Laboratory. 1994.
Jain, V. Survey of Waste Solidification Process Technologies. NUREG/CR-6666. Washington, DC:Nuclear Regulatory Commission. 2000.
Jantzen, C.M. Firstprinciples process-productmodels for vitrification ofnuclear waste: Relationship of glasscomposition to glass viscosity, resistivity, liquids temperature, and durability. Proceedings ofthe Fifth International Symposium on Nuclear Waste Management IV 93rd Annual Meeting ofthe American Ceramic Society, Cincinnati, Ohio, April 29-May 3, 1991. Ceramic TransactionsVolume 23. G.C. Wicks, D.F. Bickford, and L.R. Bunnell, eds. Westerville, OH: American CeramicSociety: 37-51. 1991.
Kim, K-D. Electrical conductivityin mixed-alkali aluminosilicate melts. Journal oftheAmerican CeramicSociety 79(9): 2,422-2,428. 1996.
Wang, E., R.K. Mohr, A.C. Buechele, and I.L. Pegg. Current density effects on the corrosion of ceramicand metallic electrode materials in waste glasses. Proceedings of the Materials Research SocietyConference. Symposium Proceedings 412. Pittsburgh, PA: Materials Research Society: 173-188.1996.
5-11
6 GLASS MELT VISCOSITY
Viscosity is a property of a liquid state and a measure ofthe resistance to shear deformation. The applicationof a shear force causes the atoms and molecules to undergo displacement with respect to each other thatcontinues with time as force continues to be applied. The viscosity of the glass forming melt and itstemperature dependence is of paramount importance in manufacturing glass. Commercial glass manufacturingprocesses require specific viscosity at each processing step. Atypical log viscositytemperature behavior ofa commercial glass is shown in figure 6-1. Glass is melted at the melting point, which has a viscosity ofapproximately 10 Pa-s (100 P), shaped between the working point and softening point, which is called theworking range, and annealed between annealing point and strain point. The glasses that have a widertemperature range between the working point and softening point viscosities are called long glasses, and theglasses that have anarrow temperature range between working point and softening point viscosities are calledshort glasses. In the nuclear waste vitrification process, glass compositions are designed to have a viscosityof less than 10 Pa-s (100 P) at 1,150 0 C, which is the upper operating limit of ajoule-heated melter. In thischapter, a briefdiscussion of glass-melt viscosity limits imposed on glass composition is followed by a reviewof studies on glass-melt viscosity, models for predicting melt viscosity using compositional information, andthe impact of melter conditions on viscosity.
6.1 MELT VISCOSITY LIMITS
The joule-heated melters for nuclear waste vitrification require the viscosity of glass melts to bebetween 2 and 10 Pa-s (20 and 100 P) at melting temperature (Hrma et al., 1994). The limits are imposedon glass melts at melting temperature to ensure the glass melt is fluid enough to homogenize and pour.
If the viscosity of the melt is below 2 Pa-s (20 P), the glass melt could increase
* Erosion of the refractory through enhanced convective currents* Volatilization of alkalis, boron, and radioactive elements such as Cs and Tc* Penetration of melt along the refractory joints* Settling of noble metals (potential for electrical shorting)
If the viscosity is greater than 10 Pa-s (100 P), the glass melt could
* Plug the pour spout during pour* Leave undissolved components in the melt* Crystallize at cold spots in the melter
6.2 VISCOSITY OF GLASS MELTS
In glasses, the addition of network modifiers, such as alkalis and alkaline earths, to glass formingoxides such as silica, tend to decrease the viscosity of the melt. Boron on the other hand, as discussed insection 4.2.5 (boron anomaly), results in anincrease in viscositywith the addition of alkalis and alkaline earthions. The viscosity is also influenced by the size of the ions and the strength of their bonds in the glassstructure as shown in figure 6-2. Glasses containing a mixture of alkali ions exhibit mixed alkali effect, a
6-1
20 1 9
18 _
16 -
(- 14 -
0myw 12_
lo0
8 -
6 -
4 -
2
Figure 6-1. Variation of the(Varshneya, 1994)
- 17
-15
Strain Point 1014.5 P -13
Annealing Point 1013 P
90)
-7Sftening Point 1 07.65 p
-5
Working Point 1 0 P 3Melting
I ~ I I I I
200 400 600 800 1000 1200 1400
Temperature (0C)
viscosity of a common soda lime silica glass with temperature
5
4
3
CL 2._
0
*5Q 1
00)
-1
-20 10 20 30
M20 (MoI%)
Figure 6-2. Viscosity of melts in binary silicate and germanate systems as a function of alkali oxide(M 2 0) concentration (Varshneya, 1994)
6-2
nonlinear behavior, with varying composition, which passes through a minimum as one alkali ion is substitutedfor the other. Figure 6-3 shows the mixed alkali effect.
The viscosity decreases as temperature increases. The temperature dependence ofthe melt viscosity(ij) of a glass is Arrhenius in nature and can be expressed as
Ea
71= r1 0 eRT(6-1)
where, h is a constant, E, is activation energy for viscous flow, R is a gas constant, and Tis the temperaturein K. Equation (6-1) can be rewritten as
log(il) = A + BT
(6-2)
whereA = logil,
and
B=_ Ea2.303R
In additionto Eq. (6-1) or (6-2), several other expressions are used to describe viscosity-temperaturebehavior (Varshneya, 1994). The most widelyused expression was described byVogel-Fulcher-Tammann(VFT), which is usually referred to as the VFT equation [Eq. (6-3)].
700
650
600 _
0. 0
E 4S50 _-o
400
3500 4 8 12 16 20
Na2O (x) Mo'%x Na2 O*(20 - x) K2O,80 SiO2
Figure 6-3. Isokom (equal viscosity) temperatures for xNa2O (20-x) K 20-80 SiO2 glasses. Log '(poise)=8.0(X); 10.0 (0); 12.0 (A); 16.0 (0) (Varshneya, 1994)
6-3
log() = -A + B (6-3)
where A, B and To are constants. If To is equal to 0, the VFT equation reverts back to the Arrhenius equation.
Glass melt viscosities in the range between 0.05 Pa-s (0.5 P) and 5 x103 Pa-s (5 x 104 P) aremeasured using the rotating viscometer. A typical experiment configuration is shown in figure 6-4. Theapparatus consists ofa temperature-controlled furnace in which a platinum-rhodium alloy crucible filled witha test sample is placed, and a probe head consisting of a Pt-Rh alloy spindle lowered into a glass melt to apredetermined height. The probe head is connected to a measuring system that applies a constant speed tothe spindle, and the resulting torque is measured. Since the cell geometry is not specifically defined, ageometric cell constant is obtained by developing torque-speed curves for standard viscosityoils and verifiedby a standard glass. Viscosity at a given temperature is determined by Eq. (6-4).
GI
N
where G, N, and -r are geometric cell constant, spindle speed, and measured torque.
(6-4)
C
Viscometer| Control Box
Spindle -
-Temp. ControlledFurnace
3lass
'rucible
Figure 6-4. Schematic of a rotating viscometer
6-4
6.3 VISCOSITY DATA AND MODELS
Viscosity is one ofthe most widely studied melt properties because of its importance in the processingand manufacture of glass. While there is an enormous amount of data on commercial glasses, this review isrestricted to viscosity data and models by Feng et al. (1990), Jantzen (1991), and Hrma et al. (1994) in usefor predicting the viscosity of nuclear waste glasses.
Feng et al. (1990) developed a viscosity-composition-temperature model based on the heat offormation. The model is based on the assumption that the bond strength, developed by Sun (table 4-1) andgiven by Eq. (6-5), between atoms in the glass is a predominant factor in controlling the compositiondependence onviscosityand otherphysical properties, such as durability. The bond strengths can be obtainedfrom known heat of formation ofthe components. The role of various oxides inthe glass can be categorizedinto network formers, such as SiO2, A12 03 , ZrO2, intermediates, and network modifiers (or breakers), suchas alkali oxides.
D(X-)2Hf (6-5)nZ
where AHfisthe standard heat offormation ofX,,O, oxide and Zis the coordination number ofX. Formationenergies were estimated using the following four rules:
(1) For SiO2, A12 0 3 , and ZrO2, which form the network of the glass, formation energies, AHj, wereobtained by Eq. (6-6).
AHj = 2AHL (6-6)
where Aj is the standard heat of formation of the pure oxide i. The factor of two is an average
ratio of binding energy and heat of formation.
(2) For network modifiers, the formation energy, AHi, was calculated from the standard heat offormation forthe pure oxides less an energyterm (E,,¢) that corresponds to mean binding energycreated by network formers as shown by Eq. (6-7).
AH = AH - E
where (6-7)
Enet =aim;
where mi is the mol fraction of glass formers in rule 1, and a, is the number of bonds formedfrom each mol of the network former i.
6-5
(3) For intermediates, the formation energy, AlI, was assumed equal to their standard heat offormation given by
AH, = AHiO (6-8)
(4) Special rule for B203. B203 is considered an excellent glass former. However, the addition ofB203 to alkali silicate glasses decreases the viscosity of the glass. This anomaly was discussedin chapter 4. To account for this anomaly, the expression shown by Eq. (6-9) is used forcalculating the contribution to the heat of formation from B203
AH =AH - 2Enet (6-9)
The total glass formation energy is given by Eq. (6-10).
AlH = E iX1 J-HI (6-10)
Feng et al. (1990) used Eq. (6-3) to represent the temperature dependence of viscosity (poise), andthe following composition dependence was used for the constants represented in Eq. (6-3).
A = ao + a,DH
B=bo + bDH
To = to + t1 DH
Regressioncoefficients were determinedbyusing 1664 datapointsfrom372glasses. Results showedno significant contributions from parameters to, t, and bo. The following values ofthe regression coefficientswere recommended for determining the viscosity of the glasses.
a. = -4.341(0.09)
al = -0.00474(0.0003)
b = -23.451(0.3)
The values in parentheses indicate the standard deviation. This model provides a reasonablerepresentation of the viscosity behavior covering a significant composition range.
Hrmaet al. (1994) conducted amultiyear statistically designed studyto characterizethe relationshipbetween composition and properties. Table 5-1 shows target composition and upper and lower limits forvarious components in the study. Viscosity was measured on 120 glasses at three to five differenttemperatures between 950 and 1,250 'C. The melt viscosity at 1,150 'C was determined by the VFTequation. Compositional dependence of viscosity at 1,150 'C was determined by the first-order mixturemodels as shown by Eq. (6-1 1).
6-6
10
In i 1,5o = E box, (6-11)
where 1150 is viscosity at 1,150 0C in Pa-s, xi and bi are mass fraction and regression coefficients forcomponent i. The estimated regression coefficients are shown in table 6-1. Hrma et al. (1994) also used asecond-order mixture model using first-order mixture model terms and several second-order terms that wereselected byusingstatistical variable selectiontechniques to improvethe relationship betweenmeltviscosity,temperature and composition. The second-order mixture model showed some improvement in predicting themelt viscosity compared to the first-order mixture model.
Based on the regression coefficients provided in the table 6-1, a component effects plot wasdeveloped centered around compositionHW-39-4 (figure 6-5). Theanalysispredicts Si0 2, A1203, andZrO2
(in order) have the strongest effects in increasing the melt viscosity while Li2O, Na2O, CaO, and B203 (inorder) havethe strongest effectin decreasingthe melt viscosity. In addition, Hrma et al. (1994) reviewed thehistorical database of 680 glasses and 160 selected glasses whoseviscosityat 1,150 0C was available. Theviscosity of 160 glasses ranged from I to 400 Pa-s. First-order mixture models predicted r2 = 0. 8998.
Jantzen (1991), developed aviscositymodel based onthenumber ofNBO inthe glass structure. Thetotal number ofNBO is a given melt composition, as calculated byEq. (4-5). Based ontheNBO calculations,an empirical relationship shown in Eq. (6-12) was developed. The relationship provided an r2 = 0.95.
10ogn150 = -1.562NB0+ 3.306 (6-12)
Table 6-1. Regression coefficients for In Ott (Hrma et al., 1994)
I xi I bi I
SiO2 8.967982
B203 -6.204318
Na2O -11.016616
Li2O -34.239274
CaO -7.466158
MgO -2.776217
Fe203 -0.036918
Al203 11.306471
ZrO2 7.433982
Others 0.9260
6-7
30H
Inas
0
LO
Zi,0
0)0(D
Va)0-
251-
20F
15 -
10
Na 2 O\
B202
Fe2 03 -
Others
Sio -- -I I I
A1203
Li2 0
\ / SiO2
2 i/03 2 30
- v .k 9 V IXB 2 03.i 2 O CaO
U2 ~ Na2 O
5
. . . .
-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16
Component Change from HW-39-4 Value
Figure 6-5. Predicted component effects on viscosity at 1,150 'C compared to the HW-39-4composition, based on the first-order mixture model using mass fractions (Hrma et al., 1994)
6-8
where q is in poise. The combination of compositional dependence and temperature yielded the relationshipshown in Eq. (6-13) with r2 = 0.976.
logq = -0.61+ 447-4 _ 1.534NBO (6-13)T(0C)
DWPF uses Eq. (6-13) to estimate melt viscosity. Jantzen's model assumes all alkalis contributesequally to the melt viscosity. Based on Hrma's experimental work, this assumption is not supported. Hrmaet al. (1994) used viscosity data from 124 glasses and Jantzen's model to predict the melt viscosity. Resultsindicated that 84 out of 126 glasses over-predicted the viscosity ofthe composition. The r2 value was 0.697.
The models reviewed in this section were developed for a specific range of glass compositions andglass components. The applicability ofthese models to other wastes as indicated by inputting Hanford WasteVitrification Waste (HWVP) glass data in the DWPF model, however, is limited. The user, however, iscautioned to perform a careful and rigorous evaluation prior to using any model for predicting melt viscosity.In addition, the reader should ensure correct units of viscosity are used during analysis because models areeither developed using poise or Pa-s. Lower temperatures (< 1,050 'C) could lead to higher viscosities thatcould result in homogenities inthe glass (incomplete melting) or plug the melter drain. If the temperature ishigher than 1,150 CC, alkalis, boron, and radionuclides, such as Cs and Tc, could volatilize.
6.4 EFFECT OF MELTER CONDITIONS ON VISCOSITY
Melter conditions have no significant effect on viscosity provided the temperature of the melter ismaintained within the operating range to ensure that the glass melt is maintained within the prescribedviscosity range of 2 to 10 Pa-s (20 to 100 P).
6.5 TANK WASTE REMEDIATION SYSTEM CONCERNS
The LAW glass composition contains asignificantlyhigher concentration of Naions than HLW glasscomposition based on the proposed process flowsheet. The LAW glass composition could have lowerviscositythan HLW glass composition, but as long as both compositions fall within the prescribed viscosityrange, there is little risk of processing problems in the melter.
6.6 SUMMARY
Viscosity is a property of a liquid state and a measure of the resistance to shear deformation. Theapplication of a shear force causes the atoms and molecules to undergo displacement with respect to eachother that continues with time as the application of force continues. The viscosity of the glass forming meltand its temperature dependence is of paramount importance in manufacturing glass. Thejoule-heated meltersfor nuclear waste vitrification require viscosity of glass melts between 2 and 10 Pa-s (20 and 100 P) atmelting temperature (Hrma et al., 1994). The limits are imposed on glass melts at melting temperature toensure the glass melt is fluid enough to homogenize and pour. If the viscosity of the melt is below 2 Pa-s(20 P), glass melt could increase erosion of refractory; volatilization of alkalis, boron, and radionuclides;penetration of melt along the refractory joints; or settling of noble metals. If the viscosity is greater than10 Pa-s (100 P), glass meltcould plug the pourspoutduringpour, have undissolved components inthe melt,
6-9
or crystallize at cold spots in the melter. The first-order mixture model predicts SiO2 , A1203, and ZrO2 (inorder) have the strongest effects in increasing the melt viscosity, while Li20, Na2O, CaO, and B203 (inorder) have the strongest effect in decreasing the melt viscosity. The temperature dependence of the meltviscosity (ij) of a glass is Arrhenius in nature. The models reviewed in this section were developed for aspecific range of glass compositions and glass components. The applicability ofthese models to other wastesas indicated by inputting Hrma's data on HWVP glasses in Jantzen's DWPF model, however, is limited.Consequently, the user is cautioned to perform a careful and rigorous evaluation prior to using anymodel forpredicting melt viscosity.
6.7 REFERENCES
Feng, X., E. Saad, and I.L. Pegg. A model for the viscosity of multicomponent glass melts. Proceedingsof the Nuclear Waste Management III. Annual Meeting of the American Ceramic Society.Ceramic Transactions Volume 9. G.B. Mellinger, ed. Westerville, OH: American Ceramic Society:457-468.1990.
Hrma, P.R., G.F. Piepel, M.J. Schweiger, D.E. Smith, D.-S. Kim, P.E. Redgate, J.D. Vienna, C.A. LoPresti,D.B. Simpson, D.K. Peeler, and M.H. Langowski. Property/Composition Relationships forHanford High-Level Waste GlassesMeltingat 1,150 0C. Volume 1. Chapters 1-11. PNL-10359.Richland, WA: Pacific Northwest National Laboratory. 1994.
Jantzen, C.M. first principles process-product models forvitrification ofnuclear waste: relationship of glasscomposition to glass viscosity, resistivity, liquidus temperature, and durability. Proceedings ofthe Fifth International Symposium on Nuclear Waste Management IV 9 3 rd Annual Meeting ofthe American Ceramic Society, Cincinnati, Ohio, April 29-May 3, 1991. Ceramic TransactionsVolume 23. G.C. Wicks, D.F. Bickford, and L.R Bunnell, eds. Westerville, OH: American CeramicSociety: 37-51. 1991.
Varshneya, A.K. Fundamentals of Inorganic Glasses. San Diego, CA: Academic Press, Inc. 1994.
6-10
7 GLASS TRANSITION TEMPERATURE
The glasstransition(ortransformation) range (Tgrange) designatesthetemperatureinterval at which a givensystem graduallytransforms as it cools from a supercooled liquid state into the glass state (Varshneya, 1994).Figure 4-1 shows a volume temperature diagram for a glass-forming melt. Glass-forming liquids, when cooledfrom location a, skip the crystallization temperature at locationb and continues to follow the path shown. Theglass transition range begins at a point at which the cooling path starts, departs from the supercooled liquidpath, and ends when the cooling profile attains a constant slope. Also shown in the figure is the effect ofcooling. The faster the glass is cooled, the higher the glass transition temperature. There is no absolutemeasure of Tg because its value is influenced bythe thermal history and rate of cooling. For most purposes,Tg is defined as the intersection of the extrapolated liquid and glassy state lines. This temperature is alsoreferred to as the fictive temperature. On a viscosity curve (figure 6-1), T. range is bounded between thestrain and annealing points (Marra et al., 1991). The annealing point has aviscosity of 1012 Pa-s (1013 P),while the strain point has a viscosity of 101i3 Pa-s (1014.5 P). Understanding of Tg of glass and itscompositional dependence is of greatimportancein commercial glass-formingprocesses. Inthe disposal ofthe vitrified HLW, Tgprovides an upper bound above which the mechanical integrity and chemical durabilityare compromised. hnthis chapter, abriefdiscussiononTglimitimposed onthe glass compositionis followedby a review of various compositional studies on Tg and models correlating composition to Tg.
7.1 GLASS TRANSITION TEMPERATURE LIMIT
The WAPS require, atthetime of shipment, the producer to certifythat, afterthe initial cool-down,the waste form temperature has not exceeded 400 0C (West Valley Nuclear Services, 1996). This productspecification was established to ensurethatthe waste is insolid formatthetime of shipment. At temperaturesabove Tg, glass is considered to be in a liquid state, and the stability of the glass could be compromised dueto nucleation and crystallization of new phases. These changes can affect chemical durability and mechanicalintegrity.
7.2 GLASS TRANSITION TEMPERATURE MEASUREMENTS
The commonly used methods for measuring Tg are differential scanning calorimetry (DSC),dilatometry, and fiber elongation (ASTM C336-71). DSC measures the heat-generation rate as a functionoftemperature. The change in the slope ofthe heat-generation rate profile as shown in figure 7-1 representsthe onset ofthe Tgrange. The Tg is defined at the intersection ofthe extrapolated lines as shown inthe figure.Dilatometry measures the thermal expansion in a sample as a function of temperature. The change in theslope of the thermal expansion profile (figure 7-2) represents the onset of the Tg range. The Tg is defined asthe intersection ofthe extrapolated lines as showninthe figure. The reproducibility of datais ±5 0C betweenDSC and dilatometrymethods as discussed in section 7.3. The fiber elongation method is based on measuringviscosity of 1012 Pa-s (1013 P) at the annealing point and a viscosity of 1013.5 Pa-s (1014.5 P) at the strainpoints. The fiber elongation method is an ASTM method widely used in the commercial glass industry todetermine annealing and strain point of glass compositions (American Society for Testing and Materials,1991). DSC and dilatometrymethods were used by WVDP to determine Tg, while the fiber elongation methodwas used by DWPF to determine Tg.
7-1
2.00
CD3
1.00
Tg 45 1 0C
o.ocIIII350 370 390 410 430 450 470 490 510 530
Temperature (0C)
Figure 7-1. Determination of transition temperature using differential scanning calorimetry (WestValley Nuclear Services, 1996)
7-2
12
10
8
E
=6
0.
cox 42
Tg =450 00
0 1 1 10 80 160 240 320 400 480 560 640
Temperature (00)
Figure 7-2. Determination of transition temperature using dilatometry (West Valley NuclearServices, 1996)
7-3
7.3 GLASS TRANSITION TEMPERATURE DATA AND MODELS
There is no absolute measure of Tg because its value is influenced by the thermal history and rateof cooling. Tg is a required product specification for the disposal of the vitrified HLW, and is stronglydependent on chemical composition (e.g., Tg of fused silica glass can be lowered from 1,200 to below 4000C bythe addition ofalkali oxides). As alkali oxides are added to the silica structure, the network connectivityis broken, which reduces the structure rigidity, and, hence, the Tg. For most vitrified HLW, T. is above 400'C (required by the specification) but below 500 'C. To obtain Tg above 500 'C requires melting attemperatures higherthan 1,150 'C, which exceeds the maximum operatingtemperature forjoule-heated Alloy690 melters. The Tg ranges for the various HLW target compositions at DWPF are shown in table 7-1.Measurements indicate no significant changes in Tg despite differences in the chemical composition. This lackof change could be attributed to the fact that for the given variabilityin compositions, the net effects of variouscomponents on Tg cancel each other. The differences inthe chemical composition (major components only)are shown in table 7-2.
EventhoughTgis awidelymeasured propertyinthe glass industry, no systematic attempts havebeenmade to establish correlation of Tg to chemical composition. For the vitrified HLW, Hrma et al. (1994)conducted amultiyear statistically designed studyto characterize the relationships between composition andproperties. Table 5-1 shows target composition and upper and the lowerlimits for various components inthestudy. The target composition, HW-39-4, is based on the proposed Hanford HLW glass composition fortheNCAW (now known as envelope B/D waste). Tg was measured on 120 glasses using dilatometry.Compositional dependence onTg was determined bythe first-order mixture models as shown by Eq. (7-1).
10Tg =Xbixi (7-1)
where Tg is glass transition temperature in 'C, xi and bi are mass fraction and regression coefficients forcomponent i, respectively. The estimated regression coefficients are shown in table 7-3. Even though r2 was
Table 7-1. Transition temperature for Defense Waste Processing Facility high-level wastecompositions using fiber elongation method (Marra et al., 1991)
Glass Composition* J Annealing Point ('C) J Strain Point ('C)
Blend 446 421
Batch 1 442 418
Batch 2 446 421
Batch 3 446 420
Batch 4 446 420
HM 460 432
PUREXt 445 421
* See table 7-2 for the compositiontPlutonium uranium extraction process
7-4
Table 7-2. Defense Waste Processing Facility projected compositions (Marra et al., 1991)
GlassComponents
wt% Blend HM PUREX* Batch 1 Batch 2 Batch 3 Batch 4
A120 3 4.16 7.15 2.99 4.88 4.63 3.44 3.43
B20 3 8.05 7.03 10.33 7.78 7.88 7.69 8.14
CaO 1.03 1.01 1.09 1.22 1.08 0.99 0.84
CuO 0.44 0.25 0.42 0.40 0.42 0.40 0.45
Fe203 10.91 7.78 13.25 12.84 11.12 11.71 11.71
K20 3.68 2.21 3.41 3.4 3.38 3.40 3.86
Li2O 4.44 4.62 3.22 4.43 4.50 4.51 4.29
MgO 1.41 1.49 1.14 1.42 1.42 1.42 1.43
MnO2 2.05 2.15 2.07 2.11 1.73 1.87 3.11
Na2 O 9.13 8.56 12.62 9.00 9.21 9.01 9.16
NiO 0.89 0.41 1.19 0.75 0.90 1.05 1.06
SiO2 51.9 55.8 46.5 50.2 52.1 52.6 50.1
TiO2 0.89 0.56 0.68 0.68 0.69 0.68 1.03
ZrO2 0.14 0.33 0.05 0.10 0.17 0.12 0.22
P~utoiium uranium extraction process
Table 7-3. Regression(Hrma et al., 1994)
coefficients for transition temperature
I xi I b, ISiO2 622.973
B20 3 584.939
Na2O 128.504
Li2O -571.109
CaO 621.517
MgO 494.391
Fe 203 427.129
A1203 544.451
ZrO2 730.071
Others 363.637
7-5
0.8822, the data showed that the first-order mixture model over-predicted lower Tg and under-predicted higherTg. Hrma et al. (1994) also used second-order mixture model using first-order mixture model terms andseveral second-order terms that were selected by using statistical variable selection techniques to improverelationship between Tg and composition. The second-order mixture model did not fit the Tg data substantiallybetter than the first-order mixture model.
Based on the regression coefficients provided in table 7-3, a component effects plot was developedcentered around the composition HW-39-4 (figure 7-3). The analysis predicts that Li2O would have a muchstronger effect in decreasing the Tg as compared to Na2O. SiO2 and ZrO2 increase the Tg, although not asstrongly as Li2O and Na2 O decrease it. CaO and B203 also increase the To, but not as much as SiO2 andZrO2. In addition, Hrma et al. (1994) reviewed historical database of 680 glasses and selected 143 of theseglasses whose Tg values were available. The Tg of 143 glasses ranged from 280-690 'C. A first-ordermixture model predicted r2 = 0.3871, which indicated that historical data were not representative of thecomposition range studied. The model reviewed in this section was developed for a specific range of glasscompositions and glass components. The applicability of these models to other wastes, however, is limited.
Table 7-4 shows the effect of redox conditions on Tg for the WVDP reference glass. The Fe2+/Fe3+ratio was changed from 0 to 1.04, and the Tg was measured using DSC and dilatometry. Results indicate thatthe Tg decreases by21 'C from 468 'C to 447 'C as the Fe2+/Fe3+ changes from a fully oxidized conditionto 1.04. This slight reduction in Tg is attributed to the increase in Fe2+ ions, which act as network modifiersand the decrease in the Fe 3+ ions, which act as network formers. Because network modifiers tend to increasethe number ofNBO sites or reduce the network connectivity, the Tg decreases. Within the normal operatingrange of Fe2+/Fe3+ of 0.10 and 0.50, no significant change in Tg was observed.
To simulate the effect ofthermal history and the cooling rate of glass in a canister, WVDP referenceglass was subj ected to various isothermal heat treatments at different temperatures and durations. In addition,glass-cooling profiles from various locations within a canister were simulated in a furnace, and the sampleswere non-isothermally heat treated to determine the effect of cooling rates on Tg. The Tg data on isothermaland nonisothermal heattreated glasses are shown in table 7-5. The sample ID WVCM70-OX-600-96 refersto WVCM70 oxidized glass heat treated at 600 'C for 96 hr. The channel number in the table refers tononisothermal temperature profiles inthe canister during cool-down. Atypical cooling profile inside aWVDPcanister, is shown in figure 7-4. The solid line refers to cooling at the surface of the canister while the dashedline refers to cooling at the center of the canister at a specific height in the canister. The isothermally heatedsamples indicate a slight increase in Tg. The samples cooled along the canister cooling curves did not showany significant change in Tg.
Literature review indicates that minor changes in composition, redox, or cooling rate do not result insignificant changes in Tg. If the Tg is close to WAPS (400 'C), the effects of redox and glass-coolinghistorycould become important.
7.4 EFFECT OF MELTER CONDITIONS ON GLASS TRANSITIONTEMPERATURE
Melter conditions such as redox show a slight decrease in Tg as the redox ratio increases as shownin tables 7-4 and 7-5 for the West Valley glasses (West Valley Nuclear Services, Inc., 1996).
7-6
520 F-u20
a
0
M
2a)0.E
2C0
C
I-'a-D
a)C-
ZrO2
510 H
SiO2 / - - CaO
-I -- ,- B2 O3500 F-Others
490 H - - MgO
\ .' I -Others - Fe203B2 03 .--
480 H
470 H
460 HSl 2
\ U201 1 1 1 1 1 1 1
\ Na2OI I I I I I. . . . . .I I I
-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16
Component wt% Change from HW-39-4 Value
Figure 7-3. Predicted component effects on transition temperature relative to the HW-39-4composition, based on the first-order mixture model using mass fractions (Hrma et al., 1994)
7-7
Table 7-4. Effect of Fe2"/Fe 3" on transition temperature (0C) (West Valley Nuclear Services, 1996)
TransitionTemperature (0C)using Differential Transition Average
Sample Scanning Temperature (0 C) TransitionIdentification Fe"M/Fe3" Calorimetry using Dilatometry Temperature (0C)
WVCM70-0 0.00 468 467 468
WVCM70-13 0.15 454 455 455
WVCM70-28 0.39 452 452 452
WVCM70-51 1.04 447 446 447
Table 7-5. Effect of isothermal and nonisothermal heat treatment on transition temperature (0 C)(West Valley Nuclear Services, 1996)
TransitionTemperature (0C) Transition Averageusing Differential Temperature (0 C) Transition
Sample Identification Scanning Calorimetry using Dilatometry Temperature (0 C)
WVCM70-OX-600-96 456 453 468
WVCM70-OX-700-96 462 462 455
WVCM70-OX-800-24 455 453 454
WVCM70R-600-96 466 465 465
WVCM70R-700-96 465 464 464
WVCM7OR-800-24 456 452 454
WVCM7I-OX-Channel-I 455 452 454
|WVCM70-OX-Channel-5 454 453 454
WVCM70-OX-Channel-6 455 453 454
WVCM70-OX-Channel-53 453 451 452
NWVM7O Channel-57 452 456 454
WVCM70-OX-Chamnel-58 451 450 451
7-8
0 750700
4 650E 600
550 -
500
450
400
350
3000 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30
Time (hr)
Figure 7-4. Typical canister profiles in the West Valley Demonstration Project (Jain and Barnes,1991)
7.5 TANK WASTE REMEDIATION SYSTEM CONCERNS
Tg is an important specification for the vitrified HLW. The target glass composition should bedesigned to meetthe specification and to accommodate variabilityin Tg due to composition variability, redox,or thermal-cooling history.
7.6 SUMMARY
The WAPS require, at the time of shipment, the producer to certify that after the initial cool-down,the waste form temperature has not exceeded 400 'C. This product specification was established to ensurethat the waste is in a solid form at the time of shipment. Tg range marks the temperature interval at whicha given system gradually transforms on cooling from a supercooled liquid state into the glass state. There isno absolute measure of'Tgbecause its value is influenced bythe thermal history and rate of cooling. For mostpurposes, Tgis defined as the intersection ofthe extrapolated liquid and glassystate lines. The commonlyusedmethods for measuring Tg are DSC, dilatometry, and fiber elongation. A CVS byHrma et al. (1994) showedthat Li20 has a much stronger effect compared to Na20 in decreasing the Tg. Si0 2 and ZrO2 increase theT., although not as strongly as Li20 and Na 2O decrease it. CaO and B203 also increase the T., but not asmuch as SiO2 and ZrQ2. Literature review indicates that minor changes in composition, redox, or cooling ratedo not show significant change in Tg. If the Tg is close to 400 'C, the effect of redox and glass-cooling historycould become important. Tgis an important specification forthe vitrified HLW. The target glass compositionshould be designed to meet the specification and to accommodate variability in Tg due to compositionvariability, redox, or thermal-cooling history.
7-9
7.7 REFERENCES
American SocietyforTesting and Materials. Standard testmethod for annealing point and strainpoint of glassby fiber elongation. C336-71: Annual Book of ASTM Standards. Volume 15.02: GeneralProducts, Chemical Specialities, and End Use Products. Philadelphia, PA: American Society forTesting and Materials. 1991.
Hrma, P.R., G.F. Piepel, M.J. Schweiger, D.E. Smith, D.-S. Kim, P.E. Redgate, J.D. Vienna, C.A. LoPresti,D.B. Simpson, D.K. Peeler, and M.H. Langowski. Property/Composition Relationships forHanfordHigh-Level WasteGlassesMeltingatl,150 0C. Volume 1. Chapters 1-11. PNL-10359.Richland, WA: Pacific Northwest National Laboratory. 1994.
Jain, V., and S.M. Barnes. Effect of glass pour cycle onthe crystallizationbehaviorinthe canistered productat the West Valley Demonstration Project. Proceedings of the Fifth International Symposium onNuclear Waste Management IV 9 3rd Annual Meeting of the American Ceramic Society,Cincinnati, Ohio, April 29-May 3, 1991. Ceramic Transactions Volume 23. G.C. Wicks, D.F.Bickford, and L.R. Bunnell, eds. Westerville, OH: American Ceramic Society: 239-250. 1991.
Marra, S.L., C.M. Jantzen, and A.A. Ramsey. DWPF glass transitiontemperatures-What they are and whythey are important. Proceedings of the Fifth International Symposium on Nuclear WasteManagement IV 93rd Annual Meeting of the American Ceramic Society, Cincinnati, Ohio, April29-May 3, 1991. Ceramic Transactions Volume 23. G.C. Wicks, D.F. Bickford, and L.R. Bunnell,eds. Westerville, OH: American Ceramic Society: 465-473. 1991.
Varshneya, A.K. Fundamentals of Inorganic Glasses. San Diego, CA: Academic Press, Inc. 1994.
West Valley Nuclear Services. Waste Qualification Report. WVNS-1 86. Revision 1. West Valley, NY:West Valley Nuclear Services. 1996.
7-10
8 LIQULDUS TEMPERATURE
Liquidus temperature (TL) is the highest temperature at which melt and primary crystalline phases canco-exist at equilibrium. At temperatures higher than the T,, no crystalline phases are present in the melt. Ifthe lowest temperature in the melter is lower than the TL, crystalline phases can precipitate and causeprocessing problems. If the crystalline phases are electronically conductive, electrical shorting could occurin the melter. Experimental vitrification studies using joule-heated melters in the late 1970s at the PacificNorthwest National Laboratory (PNNL) and at the SRL produced large amounts of insoluble crystallinephases in the melter (Jantzen, 1991). The dominant crystalline species formed at the PNNL melter wascerium oxide, while nickel-iron spinel was most dominant in the SRL melter. These observations led to thedevelopment of more robust melters and glass compositions to handle crystal formation in the melter.Therefore, the understanding ofTL and its compositional dependence is of great importance for the operationof electrically (joule) heated melters. In addition, the data from the first year of DWPF operation show thatwaste loading is limited by TL with spinel as a primary phase. Any improvement in increasing the wasteloading through better understanding of the TL could result in significant cost savings. A 1 wt% increase inDWPF glass waste loading could reduce cleanup costby$200million(Hrmaetal., 1998). Glass compositionsare designed to provide specific TL. In this chapter, a brief discussion on TL limit imposed on the glasscomposition is followed by a review of various studies on TL and models correlating TL to composition.
8.1 LIQUIDUS TEMPERATURE LIMITS
Liquidus temperature is a process property and requires that the lowest temperature in thejoule-heated melters be abovethe TLofthemelt. Injoule-heatedmelters operating at an average temperatureof 1,150 0C, the TL limit is set at 1,050 'C.
8.2 LIQUIDUS TEMPERATURE MEASUREMENTS
TL is measured by heating glass samples in a furnace and determining the highest temperature atwhich crystals can be formed in the melt. Each sample is placed in a Pt-Au or Pt- 10 percent Rh containerwith lid and heat treated for 24 hr. The use ofatemperature-gradient furnace offers a quick method to studythe entire temperature range in a single test compared to a uniform temperature furnace, which requires eachsample be heat-treated separately. In the temperature gradient furnace, however, TL measurements couldbe compromised (i) due to excessive volatilization at the hot end ofthe furnace and condensation at the coldend of the furnace, (ii) drifting of crystalline phases from the cold end toward the hot end ofthe furnace, and(iii) convective flow of glass in the sample container, which is shaped like a boat.
The heat-treatment temperature could be reached by heating the sample to the heat-treatmenttemperature or cooling the glass melttotheheattreatmenttemperature. Heating the sample is preferred overcooling because (i) melting small samples at T> TL could cause volatilization; (ii) a portion ofthe melt, whichcould be significant due to small sample size, could be lost due to capillary action between crucible and lid;and (iii) melting could destroy nuclei and prevent the sample from crystallization. After heat treatment,samples are thin-sectioned and analyzed in an optical or scanning electron microscope to evaluatecrystallization. Other than volatilization during heat treatment, TL could be affected by the presence ofinsoluble species such as RuO2, which could act as nucleating agents and redox conditions.
8-1
8.3 LIQUIDUS TEMPERATURE DATA AND MODELS
The importance of TL in ajoule-heated melter forvitrifyingHLW is clearly evident fromthetestrunsat PNNL and SRL, in which significant amounts of crystals precipitated in the melter. Since these test runs,several detailed studies have been conducted to design glass compositions to minimize crystal formation duringmelting. Crystal formation in glass melts cannot be completely eliminated, but glass compositions can bedesigned to provide minimum crystallization during melting. The TL is the main property that is used in theglass composition design process for minimizing crystal formation during melting. Even though operating amelter at temperatures higher than TL ensures minimum crystal formation in the melt, the settling of limitedsolubility species such as noble metals or formation of sulfides and metals due to uncontrolled reducingconditions in the melter could result in formation of secondary phases. The formation of these secondaryphases resulting from uncontrolled melter conditions is discussed in chapter 11.
Jantzen (1991) developed the TL model for spinels in the DWPF HLW glasses. The model is basedon the free energy offormation ofthe liquidus phases-spinel and nepheline (Na ASiO4), which are the twomain phases formed in the DWPF process. The model assumes that
* Amorphous SiO2 is the predominant solvent in the borosilicate waste glass.* Fe203 is considered as the limiting species for spinel precipitation.* A1203 is considered as the limiting species for nepheline precipitation.
Based on the TL measurements on 30 glasses, Jantzen (1991) developed the following relationshipfor calculating TL.
TL = 803.6 + 2277x r2 = 0.77 (8-1)
where ic is defined as a pseudo-equilibrium constant and is expressed as
K = - m(Fe 2 0 3) X AGfm[NiFe 2 0 4 ] (8-2)m(SiO2 ) x AGfm [SiO2 ] - m(A12 0 3 ) x AGf [NaA1SO 4 (
In Eq. (8-2), m(Fe2O3), m(SiO2), and m(A1203) are mole fractions of Fe203, SiO2, and A1203 in theglass and Gibbs free energy of formation at 1,050 'C are
AGfm [NiFe204] = -134 kcallmolAGfm [SiO2] = 156 kcallmol
AGfm [NaAlSiO4] = -360 kcallmol
DWPF routinely uses Eq. (8-1) to predict the TL, but uncertainty is significant because the model isbased on a limited number of samples. Currently, additional TL data on 53 glasses have been collected andare being analyzed to improve the TL model (Hrma et al., 1998).
Hrma et al. (1994) conducted a multiyear statistically designed study to characterize relationshipbetween composition and properties. Table 5-1 shows target composition and upper and lower limits for
8-2
various components inthe study. Liquidus temperature was measured on 1 10 glasses. Figure 8-1 summarizesthe TL of major crystalline phases identified inthe study. The TL for spinels, clinopyroxene, and Zr-containingcrystals were modeled by the first-order mixture model as shown by Eq. (8-3).
10
TL= E bx, (8-3)
where TL is Tb, xi and bi are mass fraction and regression coefficients for component i, respectively. Theestimated regression coefficients for various types of crystals are shown in table 8-1. The predicted versusmeasured TL values for clinopyroxene, spinel, and Zr-containing crystals are shown in figures 8-2, 8-3, and8-4, respectively. Overlaid on figures 8-3 and 8-4 are the measured TL for several samples containing spineland Zr-containing crystals that had TL greater than the hot-end temperature (1,100 'C). These data were notincluded in the calculation for the first-order mixture model coefficients but were used to provide informationon the extrapolative predictive capability of the model. The r2 = 0.906 for clinopyroxene, as shown intable 8-1, shows a good correlation between TL and glass composition. However, the fit for spinels was poor(r2 = 0.643) and for Zr-containing crystals, the fit could be obtained only after including three different typesof Zr-containing crystals (Zircon, sodium zirconium silicate, and ZrO2), which provided a sufficiently largedata set to perform regression analysis. The fit for Zr-containing crystals was acceptable with r2 = 0. 79.
For the six glasses with TL greaterthanthehot-end temperature for Zr-containing crystals, the modelpredicted TL to be higher than with the hot-end temperature for all six glasses. These data suggest that themodel for Zr-containing crystals could be used for extrapolative estimation of TL for glasses outside theexperimentallytested range. For 15 spinel-containing glasses with TLhigherthanthehot-endtemperature, the
.
Spinel
Cr 2 O3
Zircon
Na-Zr silicateOlivineOrthopyroxene
NephelineClinopyroxene
SiO2
CaSiO3 e
Li2SiO3
* A
0 000 Ga
0 00 CE0 EiJEh
00 0
ODi~a 0 0X XXX<
truth~~~~II
x
600 700 800 900 1000 1100Liquidus Temperature (0C)
Figure 8-1. Liquidus temperatures of major crystalline phases in compositional variability studyglasses. The quantities in parentheses at the right of the figure represent the number of sampleswith TL 2 1,110 0C (Hrma et al., 1994)
8-3
Table 8-1. Regression coefficients for liquidus temperature (Hrma et al.,1994)
Clinopyroxene Spinel Zr-containingXii b, unit b,, unit crystals bi, unit
SiO2 855.65 989.31 753.78
B203 314.72 666.42 1095.83
Na2O 38.83 3.77 74.31
Li2O -207.05 -128.77 -956.39
CaO 1372.44 1366.21 886.76
MgO 2387.62 2830.58 2458.47
Fe2O3 1506.69 2256.00 1461.04
A1203 1319.48 1735.03 1138.06
ZrO2 1844.50 928.11 4541.99
Others 1357.40 1005.56 657.99
Number of data points 29.00 28.00 22.00r2 0.91 0.64 0.79
Measured T. Range (°C) 761-969 800-1129 856-1129
I uuu I.
0a)2
2a)E
.-j-o-0
*0.a(D
0
0
0
900- 0
0 ",Y 0
80010
0Clinopyroxener2 = 0.9054r2 (ADJ) = 0.8605
800 900Measured Liquidus Temperature (0C)
1000
Figure 8-2. Predicted versus measured liquidus temperature of clinopyroxene for the first-ordermixture model (Hrma et al., 1994)
8-4
0
12000
,- 11000a)900.a)A- 1 1000WD-Ea)90
0~
0B
I-_00
I-
00 ° 00 00 01-
0
Spinel
r2 = 0.6433r2 (ADJ) = 0.4629
0
800
800 900 1000 1100
Measured Liquidus Temperature (°C)1200
Figure 8-3. Predicted versusmodel (Hrma et al., 1994)
measured liquidus temperature of spinel for the first-order mixture
I---1At It I. --- I
°1200a)
a)
0. 1100E
I-0
c, 1000
0~x
-5 900
a)
o Zircono Na-Zr Silicate* ZrO2A TL (Measured) > 111 80C
1--
0
A
0
0
0
Zr-containing crystals-2 0.7902
r2 (ADJ) = 0.6328
0 .
.
I- .
-1 II -uvv
800 900 1000 1100Measured Liquidus Temperature (0C)
1200
Figure 8-4. Predicted versus measured liquidus temperature of Zr-containing crystals for thefirst-order mixture model (Hrma et al., 1994)
8-5
model predicted TL higher than hot-end temperature for nine glasses and lower than hot-end temperature forsix glasses. The poor extrapolative prediction is attributed to the uncontrolled redox conditions during the test,which could result in varying concentrations of FeO in the sample. The presence of FeO is important in theformation of spinels.
Based on the regression coefficients provided in table 8-1, component effects plots were developedcentered around the composition HW-39-4 (figures 8-5, 8-6, and 8-7). Alkali oxides (Li2O and Na2O)decrease the liquidus of all three crystalline phases with Li2O having the strongest effect. B203 decreasesthe TL of clinopyroxene and spinel, buthas little effect onZr-containing crystals. The addition ofMgO, ZrO2,and Fe203 is most effective in increasing the formation of clinopyroxene because both Mg and Fe arecomponents of clinopyroxene. The TL for spinel was strongly increased by MgO and Fe 2O3. The effect ofMgO influence is surprising because experimental results do not indicate the presence of MgO in spinelcrystals. In Zr-containing crystals, ZrO2 showed the strongest effect in increasing the TL, closely followedby MgO. The effect of MgO was surprising. Because of low r2 values for spinel and Zr-containing crystals,these models should be used with caution.
In developing atest matrix for studying glass properties, multiple-component constraints are usuallyimposed on the glass components to ensure that the selected test matrix meets the required properties. Forthe TL, these constraints are shown in table 8-2. As shown in table 8-2, small 2 values indicate that none ofthe multicomponent constraints shows any correlation withTL. This lack of correlation clearlyindicates thatthe TL depends on the concentrations of components other than those selected as constraints. Hrma et al.(1994) also attempted to fit the TL data collected on 110 samples using a phase-equilibria TL model. Thismodel, developed byPelton et al. (1996), and it optimizes thethermodynamic properties of crystalline phasesas functions oftemperature and composition by analyzing all available phase diagrams and thermodynamicdatabases. The model lacks thermodynamic data for Fe, Ni, and Cr, which are major components ofthe spinelphase. Excluding spinels, the model predicts one or more possible primary phases with TL. The model wastested using 89 glasses The primary phase agreed with the predicted first phase in 31 glasses, and with thepredicted second or third phase in 20 glasses. The model failed to agree with any phase in 38 glasses. Theoverall fit of predicted versus measured TL was poor with 2 = 0.431. In addition, the phase equilibria modelpredicted a fairly narrow range for TL between 825 'C and 950 'C for the majority of data points; theexperimental range was between 650 'C and 1,025 'C. The phase equilibriamodel in the present form is notadequate to predict TL. According to Kim and Hrma (1994), modifications are required to improve thepredictive capability of phase-equilibria model.
The most recent TL study was completed byHrma et al. (1998) as part of Tank Focus Area activity.Hrma et al. (1998) measured TL on 53 glasses that were within and outside the current DWPF processingregion. The composition region, as shown in table 8-3, was selected using statistical experimental designmethods, and experiments were performed using a uniform temperature furnace. The SG series, whichbounds the DWPF series shown in table 8-3, is based on DWPF processing region and Hanford processingregion for HLW glass. All glasses formed an iron-containing primary phase. Nine glasses formedacmite-augite clinopyroxene phase, while the others formed spinel as the primary phase. Occasionally, spinelcrystals nucleated on the RuO2 precipitated from the glass. Spinel contained Fe, Ni, and Cr as majorcomponents and Mn, Mg, and Al as minor components. The TL span ranged from 865 'C to 1,316 'C forspinel and from 793 'C to 996 'C for the clinopyroexne phase. The study characterized the effect of glasscomponents on TL into four groups-(Cr 2 03 , NiO) >> (MgO, TiO2, A12 03 , Fe2O3 ) > (U30 8, MnO, CaO,
8-6
950
0
a)~o Q0)00*-co2 so°. aL 900Ea)c
9a)02' ac, 5 8500"a);
1L
800
-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16
Component wt% Change from HW-39-4 Value
Figure 8-5. Predicted component effects on liquidus temperature of clinopyroxene relative to theHW-39-4 composition, based on the first-order mixture model using mass fraction (Hrma et al.,1994)
8-7
10501
a)
50U)E
U)(0-.1-0
*0U)
a)a-
1000
a)(0on
coa. 950
a-U)
02
I
MgO
Al 203/ /Fe203
Na20 /
0 LiO / >_CaO
B203 SO
Other OthersOthers 3_ _ ZrO2
SiO2 … Al20
'7'7 .~~~~~~~~~B 203
Fe2O3 / LF e 2°3 I \ N a2O
900
8501.
-12 -10 -8 -6 -4 -2 0 2 4 6 8 11 12 14 16
Component wt% Change from HW-39-4 Value
Figure 8-6. Predicted component effects on liquidus temperature of spinal relative to the HW-39-4composition, based on the first-order mixture model using mass fraction (Hrma et al., 1994)
8-8
1,200 / ZrO2
1,100 _
a) 0C 0
an 2\'
= 0)
- '
o-i-LL5N
h-OaQ iR
//
1,000 F /
900 _
/ , /MgO
Si0 2 ~~Na2O - . Li I0 0 e ~ 1 0SiO2 ----------- - __ _ 203
B203 ..- Mg_.- SiO20 ers2
re2u3 / '**N./ Li2 F _/ Na20
800 _
ZrO2 /700 I I I I I I I I I I
-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16Component wt% Change from HW-39-4 Value
Figure 8-7. Predicted component effects on liquidus temperature of Zr-containing crystals relativeto the HW-39-4 composition, based on the first-order mixture model using mass fraction (Hrmaet al., 1994)
8-9
Table 8-2. Multicomponent crystallinity constraints and their correlation with liquidus temperature(Kim and Hrma, 1994)
[ Sum of Oxide (Mol fraction) I rA
CaO + MgO = 0.10 0.036
A1203 + ZrO 2 = 0.18 0.000
CaO + MgO + ZrO 2 = 0.18 0.314
Fe203+ A12 03 + ZrO2 + Others = 0.24 0.058
L ~ ~ A1203/SiO2 = 0.33 0.078
Table 8-3. Composition regions (in mass fractions of components) for Defense Waste ProcessingFacility and SP and SG glasses (Hrma et al., 1998)
Defense Waste ProcessingFacility Region SG Glasses SP Glasses
Oxide* Low High Low High Low High
SiO2 0.491 0.551 0.430 0.590 0.380 0.600
B203 0.068 0.075 0.050 0.100 0.000 0.120
A120 3 0.024 0.055 0.025 0.080 0.040 0.160
Li2O 0.043 0.050 0.030 0.060 0.000 0.030
Na2O 0.078 0.106 0.060 0.110 0.080 0.200
K20 0.021 0.026 0.015 0.038 0.000 0.000
MgO 0.013 0.021 0.005 0.025 0.004 0.060
CaO 0.007 0.013 0.003 0.020 0.000 0.000
MnO 0.011 0.028 0.010 0.030 0.000 0.040
NiO 0.001 0.012 0.001 0.020 0.000 0.030
Fe2 03 0.085 0.124 0.060 0.150 0.060 0.150
Cr203 0.000 0.002 0.001 0.003 0.000 0.012
TiO2 0.002 0.004 0.002 0.006 0.000 0.000
U308 0.008 0.050 0.000 0.055 0.000 0.000
Others1 - 0.000 0.000 0.045 0.070
*SG glasses contained RuO2 in mass fraction 0.0009. SP glasses contained RuO2 in mass fraction 0.0003. The estimated RuO2
mass fraction for the DWPF composition region is I x 10-5 to 0.0015.tOthers for SP glasses were a mixture of 22 minor components with ZrO2, Nd2O3, La2O3 , CdO, MoO3, F, and S03 > 3 wt%/o ofthe mix.
8-10
B203, SiO2) > (K2O, Na2O, Li2O). Cr and Ni strongly increase the TL, while alkali oxides decrease the TL.
Mg, Ti, Al, and Fe moderately increase the TL, and U, Mn, Ca, B, and Si have little or no effect on TL.
The liquidus temperatures for the spinel phase observed in the SG series were analyzed usingEq. (8-3). The data set was then combined with another TL study by Mika et al. (1997) that consisted of 33glasses (composition range SP shown in table 8-1) and reanalyzed. Table 8-4 shows the regressioncoefficients obtained fromtheindividual and combined data set. The SP studyvaried one component at atimearound the baseline glass. Except for MgO, MnO, and Cr203, the regression coefficients showed a goodagreement between data sets. The difference of these three components is attributed to the differences inthe composition range not covered by the studies. The TL regression coefficients using first-order mixturemodel obtained by Hrma et al. (1994) as shown in table 8-2, are significantly different from the Hrma et al.(1998) or Mika et al. (1997) study (shown in table 8-4). Because the first-order mixture models are validwithin a specified composition range, the differences in the regression coefficients between studies areexpected.
The application of TL models discussed in this section is limited because the models are based on afew glass compositions. The TL depends not only on concentrations of its own components but also on
Table 8-4. Regression coefficients for liquidus temperature (Hrma et al., 1998)
I [x SG Glass, b, SP Glass, A SG and SP Glasses, b,
A1203 2,678 3,307 2,897
B203 453 395 396
CaO 2,033 N/A 1,781
Cr203 30,313 18,864 21,553
Fe203 2,643 2,644 2,692
K20 -1,152 N/A -952
Li 2O -1,504 -1,470 -1,367
MgO 4,839 2,827 3,823
MnO 873 1,870 1,277
Na2O -1,673 -1,826 -1,734
NiO 9,959 8,210 9,652
SiO2 981 834 997
TiO2 3,554 N/A 3,620
U308 1,546 N/A 1,622
Others N/A 4,419 3,492
r2 0.95 0.94 0.93
adj r2 0.93 0.87 0.92
8-11
concentration of other components in the glass composition, and the effect of a component could be quitedifferent for different crystalline phases. Therefore, the prediction capability of present models is limited tothe range of composition studied and crystalline phases observed.
8.4 EFFECT OF MELTER CONDITIONS ON LIQUIDUS TEMPERATURE
Even though TL is defined by the glass composition, the processing conditions in the melter couldinfluence the formation of secondary phases. These secondary phases could occur because of extremelyreducing conditions inthe melt, which could precipitate highlyconductive phases, such as NiS, CuS, or FeS,or because of phase separationin the meltifthe concentration ofthe components exceeds their solubilitylimit(e.g., immiscible sodium sulfate phase could form if SO3 concentrationis greaterthan 0.25 wt% in glass). Inaddition, limited solubility species such as noble metals and refractory corrosion products could act asnucleating sites for the formation and growth of crystals.
8.5 TANK WASTE REMEDIATION SYSTEM CONCERNS
At Hanford, both LAW and HLWjoule-heated melters will be operated at 1,150 'C. The TL shouldbe below 1,050 'C. The formation of spinel and clinopyroxene crystals in HLW glass composition is moreprobable compared to LAW glass composition because HLW contains significant concentrations ofcomponents such as Fe, Ni, Mn, and Cr, along with Al, Si and Na, which are responsible for spinel andclinopyroxene formation, respectively. In the LAW glass composition, a small concentration of spinel crystalsand a higher amount of clinopyroxene crystals are expected.
8.6 SUMMARY
TL is the highest temperature at which melt and primary crystalline phases can coexist at equilibrium.At temperatures higher than the TL, no crystalline phases are present in the melt. If the TL of the melt ishigher than the lowest temperature in the melter, crystalline phases can precipitate, perhaps causingprocessing problems. If the crystalline phases are electronically conductive, electrical shorting could occurinthe melter. Experimental vitrification studies usingjoule-heated melters inthe late 1970s at PNNL and SRLproduced large amounts of insoluble crystalline phases in the melter. Liquidus temperature is a processproperty and requires that the lowest temperature in the joule-heated melters be higher than the TL of themelt. Injoule-heated melters operating at an average temperature of 1,150 'C, the TL limit is set at 1,050 'C.The TL models-SRL free-energy hydration model, PNNL first-order mixture model, phase-equilibria TL
model, and process constraint model-are based on limited site-specific glass compositions. The TL dependsnot only on concentrations of its own components but also on concentration of other components in the glasscomposition. Also, the effect of a component could be quite different for different crystalline phases.Therefore, the prediction capability of present models is limited to the range ofthe composition studied andcrystalline phases observed. At Hanford, both LAW and HLW joule-heated melters will be operated at1,150 'C, the TL should be below 1,050 'C. The formation of spinel and clinopyroxene crystals in HLW glasscomposition is more probable compared to LAW glass composition because HLW contains significantconcentrations of components such as Fe, Ni, Mn, and Cr, along with Al, Si and Na, which are responsiblefor spinel and clinopyroxene formation, respectively. In the LAW glass composition, a small concentrationof spinel crystals and a higher amount of clinopyroxene crystals are expected.
8-12
8.7 REFERENCES
Hrma, P.R, G.F. Piepel, M.J. Schweiger, D.E. Smith, D.-S. Kim, P.E. Redgate, J.D. Vienna, C.A. LoPresti,D.B. Simpson, D.K Peeler, and M.H. Langowski. Property/Composition Relationships forHanford High-Level Waste GlassesMeltingatl,150 C. Volume 1. Chapters 1-11. PNL-10359.
Richland, WA: Pacific Northwest National Laboratory. 1994.
Hrma, P.R, J.D. Vienna, M. Mika, and J.V. Crum. Liquidus Temperature Data for DWPF Glass.PNNL-1 1790. Richland, WA: Pacific Northwest National Laboratory. 1998.
Kim, D.-S and P. Hrma. Models for liquidus temperature of nuclear waste glasses. Proceedings of theEnvironmental and Waste Management Issues in the Ceramic Industry II Symposium. 9 6M
Annual Meeting of the American Ceramic Society, Indianapolis, Indiana, April 25-27, 1994.Ceramic Transactions Volume 45. D. Bickford, S. Bates, V. Jain, and G. Smith, eds. Westerville,OH: American Ceramic Society 327-337. 1994.
Mika, M., M.J. Schweiger, J.D. Vienna, and P. Hrma. Liquidus. Temperature of spinel precipitatinghigh-level waste glasses. Scientific Basis for Nuclear Waste Management XX SymposiumProceedings 465. Pittsburgh, PA: Materials Research Society: 71-78. 1997.
Jantzen, C.M. Firstprinciples process-productmodels forvitrification of nuclear waste: Relationship of glasscomposition to glass viscosity, resistivity, liquidus temperature, and durability. Proceedings ofthe Fifth International Symposium on Nuclear Waste Management IV 93rd Annual Meeting ofthe American Ceramic Society, Cincinnati, Ohio, April 29-May 3, 1991. G.C. Wicks, D.F.Bickford, and L.R. Bunnell, eds. Ceramic Transactions Volume 23. Westerville, OH: AmericanCeramic Society: 37-51. 1991.
Pelton, A.D., P. Wu, G. Eriksson, and S. Degtiarev. Development of Models and Software for LiquidusTemperatures of Glasses of HWVP Products. PNNL-11037, UC-810. Richland, WA:Pacific Northwest National Laboratory. 1996.
8-13
9 REDOX
Redox (reduction-oxidation) reactions involve a transfer of one or more electrons among the differentoxidation states ofa multivalent element or from one multivalentelementto anotherwithinachemical system.Elements that can coexist in multiple valence states are called redox couples, for example, an Fe redox couplecan exist as either Fe2+/Fe3+ or Fe0fFe2', depending on the available oxygen in the system. Knowledge ofredox couples participating in chemical reactions in glass melts and the temperature and compositionaldependence of redox reactions during vitrification of HLW injoule-heated melters is of great importance.
In this chapter, a brief discussion on redox limits is followed by a discussion ofthe thermodynamic basis forredox reactions and various variables that affect redox reactions in glass melts. In addition, redox forecastingand control strategies adopted by WVDP and DWPF to predict redox response from the HLW slurrycompositions are presented.
9.1 REDOX LIMITS
The redox state in a melter is probably the most important process control parameter. A redoxresponse within a specified range is necessaryto avoid process problems that could lead to permanent melterdamage. Redox couples that could influence the redox behavior of the nuclear waste glasses are shown intable 9-1. Because Fe is abundant in most wastes and oxidation states can be easily measured, the redoxresponse in waste vitrification systems is determined by the Fe2+/Fe3' ratio. The influence of some minorwaste components, such as Cu and Mn, cannot be ignored. A glass melt is defined as extremely reducing ifthe Fe2+/Fe3" ratio is greater than 1. Under such conditions, sufficient conductive metals and metal sulfidescould accumulate and short-circuitthe melter. If the Fe2"/Fe" ratio is less than 0. 01, a glass melt is consideredto be extremely oxidizing. Under extremely oxidizing conditions, foaming is observed in the melter. Foamcreates an insulating layer of gas bubbles between the cold cap and the melt, disrupting the thermal gradientsin the melter. If controlled at its onset, foaming typically slows the glass production rate and is not an issue.Left uncontrolled, however, foaming conditions could result in melter shutdown. A much narrower boundbetween 0. 01 and 0.5 is acceptable for processing to avoid foaming or precipitation of metal sulfides Fe2+/Fe3"in the glass.
9.2 THERMODYNAMICS OF REDOX REACTIONS
In simplified form, redox equilibrium of amultivalent element (M) in aglass melt can be representedas
M~It)+ n 02-it *-_-- M(m-n)+ +O (9-1M ) + 2 °(melt) (mclt) 4 2(gas) (9
whereAt+ and M"-*)+ represent oxidized and reduced species ofthe multivalent element M, n is the numberof electrons transferred during redox reaction, 02 is the oxygen released (the amount of which depends onthe furnace atmosphere and/or oxidizing and reducing agents present in the feed materials), and 02- is thebasicity (or oxygen ion activity) ofthe glass meltthat depends on the glass composition. Basicity for a givencomposition is constant.
9-1
Table 9-1. Redox couples in nuclear waste glasses
Reference Electrode Potential for Savannah Enthalpy of Reduction (AH) for SavannahRiver Laboratory (SRL-131) Glass at River Laboratory (SRL-131) Glass
Redox Couple 1,150 0 C (Schreiber and Sampson, 1996)* kJ/mol (kcal/mob) (Schreiber et aL, 199 0 )tAU3+/Auo >3.00 _
Pdf+/Pd0 +2.30 _Ni3+/Ni2+ +1.70 _
Eh 3 +/Rh° +1.40 _
Co'/Co2+ +1.40 _
_N__3+/_N_2+_4 +0.80 54 ± 13 (13 ± 3)
Ag'+/Ago +0.50 _Se6+/Se 4 + +0.20
Se4+/Se0 +0.20
Ce4/Ce' -0.10 46 ± 17 (11 ± 4)Cr6+/Cr3 + -0.30 42 ± 13 (10 ± 3)
SbI+/Sb3+ -0.30 _
Cu2+/Cu'+ -0.80 176 ± 42 (42 ± 10)U6+l/U5',+ -1.50 172 ± 54 (41 ± 13)Fe3+/Fe2 + (10 percent) -2.00 243 ± 105 (58 I 25)As5/As3+ -1.70 _v5+/v4+ -1.90 172 42 (41 10)U5+/U4+ -2.20 _
Cu'+/Cu0 -3.30 _Crl+/Cr -3.40 243 29 (58 7)
Mo"+/Mo5 * -3.80 _V4+/N3+ -4.00 155 67 (37 16)Eu3+/Eu 2 + -4.30 172 ± 59 (41 ± 14)Ti4+/Ti3 + -5.00 147 ± 21 (35 ± 05)Ni2+/Ni0 -5.30 343 ± 63 (82 ± 15)Sn4e/Sn2+ -5.50
Se0 /Se2 - -5.70
Co2+/Co0 -6.00 _
Fe2+/Fe0 -6.30 335 ± 42 (80 ±10)Mo,+/Mo0 -15.80 _S6+/S2- --19.20 787 ± 0 (188 ± 0)
*Schreiber, H.D., and E.V. Sampson, Jr. A corrosion model for metals in molten slags. Proceedings of the Corrosion of Materialsby Molten Glass Symposium at the 98 1h Annual Meeting of the American Ceramic Society, Indianapolis, Indiana. CeramicTransactions Volume 78. G. Pecoraro, J.C. Maira, and J.T. Wenzel, eds. Westerville, OH: American Ceramic Society: 3-15. 1996.tSchreiber, H.D., C.W. Schreiber, M.W. Riethmiller, and J.S. Downey. The effect of temperature of the redox constraints for theprocessing of high-level nuclear waste into a glass waste form. Proceedings of the Materials Research Society Symposium. Pittsburgh,PA: Materials Research Society 176: 419-426. 1990.
9-2
The equilibrium constant K(T, P) for Eq. (9-1) at a given temperature (7) and pressure (P) can be writtenas
K(T,P) = [M n+][02-]n/2 (9-2)
where fl 2 is oxygen fugacity. The brackets indicate the activities of the species. For a given glasscomposition, the oxygenion activity is constant and canbe assumed as unity. Equation (9-2) can be reducedto
K(T, P) = [M(mn)+ ] 24 (9-3)[MM+
Equation (9-3) can be rewritten as
-logJO2 =-l [M4lg +E (9-4)
In Eq. (9-4), as long asMis sufficiently dilute the activity ratio can be replaced bythe concentrationratio, can be easilymeasured. A plot of -log0AO 2) andl"-n)+/Ar+ concentration ratio will produce a straight
line of slope 4 and intercept E'. The intercept E' can be related to the relative reduction potential for a givenn
couple in a given glass. The more negative the value of E', the easier forMl"n)+ to undergo oxidation to Me.These plots are referred to as electrochemical series diagrams.
If two or more redox components, such as Fe and Mn, are simultaneously present in the melt, theredox equilibrium is attained by internal redox reaction via electron exchange as shown in Eq. (9-5):
rM.It) + n ÷(-m+ > rM,(melt) + nQI(qM. t ) (9 5)
where rand n are the number of electrons transferred by multivalent elementsMand Q, respectively, in themelt. The degree of two redox couples interaction can be estimated by their relative reduction potentials.
The relationship shown in Eq. (9-4) provides a convenient way of analyzing redox data and is widelyused in nuclear waste vitrification processes to study redox behavior. Schreiber and Hockman (1987)developed an electrochemical series ofredox couples in SRL-I 31 (simulated DWPF glass composition) andWV-205 (simulated WVDP glass composition) glasses. Figure 9-1 shows the relationship between -logA02 ) and M"-)+1/vP for various multivalent ions in a simulated glass composition. In figure 9-2, the samedata are replotted as - logj(02 ) and percentage of redox couple in reduced states. The importance ofredoxin HLW vitrification is clearly evident from this figure. Key regions that determine the processing range forHLW slurries are marked. Even though Fe precipitation is not observed until the melt attains an oxygenfugacity of 10-9, the upper limit for acceptable redox range is constrained to an oxygen fugacity of I 0-7 . Thisoxygen fugacity corresponds to an Fe2+/Fe3' ratio of 0.5. This lowering ofthe upper limit for acceptable redoxrange is attributed to the presence of sulfur in the HLW slurries. Sulfur establishes a S6+/S2 - redox couple
9-3
Percentage of Redox Couple in Reduced State1 5 10 20 30 50 70 80 90 95 99
N
0)cs1r
log (Reduced )/(Oxidized)
Figure 9-1. Redox chemistry in WV-205 melt at 1,150 'C superimposed on the SRL-131 meltsystem. Solid lines are SRL131 at 1,150 'C (broken solid lines labeled 5 and 10 percent are for5 wt% Fe and 10 wt% Fe in SRL-131); dashed lines with data points are WV-205 at 1,150 'C(Schreiber and Hockman, 1987).
in the system and results in the formation of electrically conductive sulfide phases such as FeS and NiS.These phases could electrically short the melter.
If the oxygen ftigacity is less than 10-2, redox couples such as Ce, Mn and Cr start influencing theredox response in the system causing formation of foam in the glass melt. Therefore, the redox response forprocessing nuclear waste glasses has been bounded between the oxygen fugacity of 1 0-4 and 10-1, whichcorresponds to Fe2+/Fe3+ ratios of 0.01 and 0.5.
From Eq. (9-1), redox equilibrium in glass melts depends on
0
0
Oxygen ion activityEquilibrium constant [K(T, P)], which is a function of the Gibbs free energy and temperatureRedox couple activityOxygen [021 activity
9.2.1 Oxygen Ion Activity
Oxygen ion activity is also referred to as basicity of the glass melt and is controlled by the basecomposition ofthe glass. More oxide ions are produced in the melt as additional network modifiers such asalkali oxides are added into a system of network formers such as B203 and SiO2. As the concentration ofalkali oxides increases in the glass, the basicity of the glass increases. According to Eq. (9-1), if theequilibrium constant K(T, P) is assumed to be independent of base composition andj0 2 ) is assumed to beconstant, then higher oxygen ion activity, [021-], should favor a lower oxidation (reduced) state. However,
9-4
Ti, 4-3\ Eu, 3-2
15 ,Cr3214 -/ r - Fe2-0
13V, 4-3
13- MO,5-0 .....12 _ <.....u -. ...
11 _ . - [Ni,2-0,Co,2-0,Cu,1-0,Sn,4-2] ,.-a
-) 10 _o-- 2CD , +CD
9 ~~~~~~~~~~~~~~~~~C
CG c)2- -7-
2o - ------ Cu 21 . ...... .. \ .j< CD(D3i, -8 -~~~~~~~~~~~~~~~ CD
7~~~~~~~~~~~~~
1m SC,65-
becoes oreoxiizin. Nte hattheRedox Couple in Reduced Sate an Goie/rdue 3aioista2oCr 7 - U, 5- -U, 6-5~~~~~~~Ni,3-
Fgreddoiie 9-2 i. Theitibuio ofparedoxpastates reoflvarious multgivalngthfathtte eleensin libriu constainin
model ucur o nuclear waste at115 lCassa fucpstionso the fimpos coxygexand fugeacity deermibetiando
canbesmmarze 1987)suis odce yScrie n okmn(97 n Shebre l
exmperiental data (Phaulian Dofgas 1965) for a seriss. of r alal siicatx atointe melts th0x~Owee MrefertLiNaoKioashwinfgr9-,indicate that, as the alaicontentrto of Finrae the glass, ih etbcms mrreuincrecaused the mlbecitome more oxidizing Note bcthatf the redox rtio iesuplotsged ast ane odizd/edce ratnniour insea fof alkl
Thssrctrloiuceraate glasses copoiton iso fairl complex,-20 and the RL13 exacte idectersmination of
cnbehsmavirdsized diferomce the copsituiens conduthed bylass Schreiber and Hockman (1987) adcheiretave
compresuthed siredox behaviorof Wprobande DWPFterglasses. Forf givenreox actio in thevu melttemdata
in the glass.
9-5
16
12-
coa)U-
Li
4
0 I0 20 40 60
M2 0 (MoI%)
Figure 9-3. Effect of composition of alkali (M) silicate melts on the concentration ratio Fe3"IFe2"for melts equilibrated with air at atmospheric pressure and 1,400 'C; total Fe in the melt is lessthan 0.5 percent (Paul and Douglas, 1965)
9.2.2 Effect of Temperature on Redox Equilibrium
The redox equilibrium in glass generallymoves toward the reduced side with increasing temperature.The temperature dependence ofredox equilibrium in a melt at constant oxygen fugacity can be described byapproximation of the Clausius-Clapeyron equation (Schreiber et al., 1986a)
log [M(m-)+ ] H (9-6)[M m + ] 2.303RT
where AHl is related to the enthalpy change forthe reduction reactionin Eq. (9-1). Equation (9-6) is only anapproximation because the equilibrium constant is replaced by the redox ratio, and oxygen ion activity isincluded in constant B and is assumed to be temperature independent.
Schreiber et al. (1 986a, b; 1990,1993) studied the effect oftemperature on the redox equilibrium forthe DWPF and WVDP glass compositions. Figure 9-4 shows a plot of the redox ratio (Fe2"/Fe 3) versus I /Tforthe SRS-1 31 glass containing 1 and 1 0 percent Fe. As the temperature increases, the redox equilibria shiftto amore reducing state. Figure 9-4 also indicates that, for a given oxygen fugacity, the enthalpy of reductionis endothermic, with AH= 189 ki/mol (45 kcallmol) and 306 kJ/mol (73 kcallmol) for 10 percent Fe203 and1 percent Fe203, respectively, in glass. Note that the enthalpy of reduction for 10 percent Fe2 ,03 is lower
9-6
T (0C)1150 1050 950 850
E ° I lco
SRL-131
0
° 1.0 ~i :
co -0.5;7 1 wFe',, AH =45 kcal/mol
LL 0
- AH = 73 kcal/mol 1t% Fe
0
-2.50.0007 0.0008 0.0009
1/T (-K-1 )
Figure 9-4. The temperature dependence of the Fe3"-Fe 2" equilibrium in melt composition(SRL-131) (Schreiber, 1986).
compared to the enthalpy of reduction for 1 percent Fe2O3 and is attributed to the basicity of the melt.Average enthalpy of reduction for Fe3+ to Fe2 and Fe2+to Fe0 reduction, as shown in table 9-1, for 10 percentFe2O3 in SRL-131 glass is 244 kJ/mol (58 kcal/mol) and 336 kJ/mol (80 kcal/mol), respectively. The enthalpyof reduction for otherredox couples, as shownintable 9-1, ranges from 42 to 787 kJ/mol (Schreiber, 1990).
Figure 9-5 shows the effect of increasing temperature on the redox equilibrium for the WVDPglasses (Schreiber et al., 1993). The straight lines relating to -logJ(0 2 ) and Fe2+/Fe3+ and -log0fO 2 ) andFe0/Fe2 ' are the best fit lines with theoretical slopes of 4 and 2, respectively, as shown by Eq. (9-4).
For a given oxygen fugacity, the data indicate that as the temperature increases, the melt becomesmore reducing. In addition, when compared with the DWPF redox data on SRL-I 31 glass shown in table 9-1,the redox relationships are similar. The relative reduction potentials, E' in Eq. (9-4), were -2.0 and -6.5for Fe2P/Fe3+ and Fe0/Fe2 ', respectively. In addition, the enthalpies of reduction were 150 and 290 kJ/mol(36 and 69 kcal/mol) for the reduction of Fe3+to Fe2" and Fe2 'to Fe0 , respectively. Slight differences in theenthalpy of reduction between DWPF and WVDP glass melt are attributed to compositional differences.
9.2.3 Effect of Redox Couple Activity
The redox equilibrium established by a multivalent element remains constant, as long as theconcentration ofthe multivalent element dissolved inthe glass meltis dilute. Schreiber (1986) has notedthatthe redox equilibrium is invariant provided the concentration of the multivalent element does not exceed2-4 wtv/ointhemelt. If the multivalentelementbecomes amajor component ofthe glass melt, the multivalentelement also contributes to the changes in the redox equilibrium because of the changes in the melt basicity.
9-7
[(Fe2+)/(Fe3 +)] or [(Fe0)/(Fe2 +)]
0.01 0.05 0.1 0.2 0.6 1 5 10
14 0 ... ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.....14 - 0~~~~~- -
.0012 -
cJ0
0
'6-
4
2-
-2.5 -2.0 -1.5 -1.0 -0.5 0 0.5 1.0 1.5
log [(Fe2 +)/(Fe3 +)J or [(Fe0 )/(Fe 2 +)
Figure 9-5. The dependence of the redox state of iron on both the imposed oxygen fugacity and the
melt temperature in West Valley glass. Circles represent results obtained at 1,250 'C, while
squares represent those at 1,050 'C. Solid symbols are the [Fe"]J/Fe"]J ratios, and open symbols
are the [Fe'J/Fe`~] ratios. Redox relations at 1,250, 1,150, and 1,050 'C are shown by dashed,
solid, and dotted lines, respectively (Schreiber et al., 1993).
Melt basicity, which is a function of oxygen ion activity, increases as the melt becomes more oxidizing. In
most of the borosilicate-based nuclear waste glasses, concentration of Fe is sufficient to be considered a
major element in the melt. The effect of increasing Fe concentration is clearly shown in figures 9-1 and 9-4
for the SRS-131 glass composition.
As indicated in Eq. (9-5), if two or more multivalent elements are simultaneously present in a melt,
the redox couple may interact by intemnal electron exchange. Schreiber et al. (1 993) studied the effect of other
redox couples on Fe"~/Fe"~ in the WYVDP glass melt. Multivalent oxides present the WvVDP HLW include
Fe (1 2 wt% Fe2O3), Ni (0. 25 wt% MiO), Mn (I wt% MnO2), Ce (0. 16 wt%/ CeO 2 ), Cr (0. 14 wt% Cr 2O3)and Cu (0.06 wt% CuO). Individual experimental studies with 1 wt%Cr 2O 3, 1 wt% NiO, 1 wt% CuO, I wt%
CeO 2, and 4 wt% MnO. additions in the WVDP glass melt showed that none of the added metals affected
the redox limits established by Fe 2 /Fe3" between 0. 01 and 0. 5. However, the DWPF glass melts that contain
similar redox couples as WvVDP glasses showed a significant Cu0 precipitation in glass melts containing more
than 0.7 wt % CuO within their redox processing range (Ramsey et al., 1994).
9.2.4 Oxygen Activity in Glass Melts
The oxygen activity, 102], in the glass melt depends on the melter atmosphere or the oxidizing and
reducing agents present inthe feed materials. For simplicity, the activity of oxygen inthe melter atmosphere
is generally replaced bythe partial pressure of oxygen. This assumption is permissible only at low pressures.
At high pressures, deviations could be significant, especially in cases where the gas reacts with the melt.
9-8
In redox reactions, even though molecular oxygen is present in the gas above the melt, only molecularoxygen dissolved within the melt can react with multivalent elementM The dissolved oxygen in the meltestablishes an equilibrium with the gaseous oxygen. Redox equilibrium is never completely achieved inmelters. The participation of molecular oxygen canbe shown by separation ofEq. (9-1) into its half reactions
Mn++ ne7 M<-n+ 9-(melt) (melt) (97a)
2 (melt) 4 2(meltgas) + ne (9-7b)
[02(melt)] = kf (0 2 ) (9-8)
where brackets indicatethe activities ofthe species. The [02(melt)] isthe oxygen activityinthe glass melt. TheHenry's Law constant, k, is dependent on the melt composition and temperature and is expected to be obeyedfor relatively low concentrations of dissolved oxygen in the melts. If, in a melter containing multivalentelements, melt conditions such as temperature or atmosphere (oxygen fugacity) are changed, the system, inaccordancewithLeChatelier's principle, willreadjustto achieve anew redox equilibrium. Such readjustmenton perturbation in the melt, however, does not happen instantaneously because molecular oxygen has todiffuse into or out ofthe meltto attain new equilibrium. Because oxygen is a major component of the HLWglass melts, its kinetic behavior could control the final redox equilibrium in the melt or determine the timerequired to reach steady state.
HLW melter feed, which consists of nitrates, nitrites, sulfates, carbonates, and hydroxides, and otherorganic components such as sugar or formic acid, undergoes chemical decomposition reactions at differenttemperatures. The decomposition reactions start as low as 100 0C and continue up to 1,000 'C. Ryan (1994)has tabulated potential reactions for major components that could occurinthe melter. As these decompositionreactions occur in the cold cap (a crusty region on the surface of the melt where most of the decompositionreactions occur), reaction products consisting of metal oxides move into the glass melt while gases arereleased at the surface of the melt. This complex mixture of reaction gases and in-flow of air from melterin-leakage determines the partial pressure of oxygen in the melter. Depending on the size and type of themelter, the time required to attain steady-state redox could range from a few hours to days.
Because the HLW feeds typically contain significant concentrations of oxidizing agents such asnitrates and nitrites, the selection of the proper reducing agent is important. The selection of the type andamount of reducing agent depends not onlyonits decompositiontemperature but also on the decompositiontemperature of oxidizing agents. The effectiveness of the reducing agent could be compromised if thedifference between decomposition temperatures is large. Starch, sugar, formates, urea, or carbon black couldreact differently in the same feed. Table 9-2 shows the effect of various reducing agents on the Fe2"/Fe 3"ratio (Bickford and Diemer, 1986).
9-9
Table 9-2. Effect of various redox compounds in closed crucible tests (Bickford and Diemer, 1986)
Sample Fe2"/Fe 3" Log (pO2 ) at 1,150 'C
90 SRL165frift +0Fe 2 0 3 +
4 NaCOOH + 1 Ni(NO3)2.6H20 0.06 -3.20
4 NaCOOH + I NaNO3 0.09 -3.72
4 NaCOOH + 2 MnO2 0.18 -5.02
0.2 C 0.28 -5.79
1.0 phenyl boric acid (PBA) 0.37 -6.29
4 NaCOOH + I Mn(COOH)2 + 0.42 -6.501 MnO2 + I Ni (NO3) 2.6 H20
4 Mn(COOH)2 0.51 -6.83
2 NaCOOH 0.54 -6.93
5 Fe(COOH)3 0.70 -7.38
6 Mn(COOH)2 0.71 -7.41
2 PBA 1.11 -8.18
5 NaCOOH 1.40 -8.58
1 C 1.45 -8.65
5 PBA+ 1 CaCO 3 2.00 -9.20
5 PBA (soot in crucible after test) 100.00 -16.00
9.3 REDOX STRATEGY ATVITRIFICATION PLANTS
HIGH-LEVEL RADIOACTIVE WASTE
The presence of multiple redox couples as shown in table 9-1, along with a number of reducing andoxidizing agents in the HLW, makes redox control in a HLW melter complex and challenging. Even thoughmost ofthe HLW stored at West Valley, Savannah River, and Hanford sites originated from the reprocessingof fuel assemblies by the plutonium uranium extraction (PUREX) process, their waste characteristics aredifferent because the fuel assemblies reprocessed at each site were different. While it is accepted by thescientific community that redox, as measured by Fe2"/Fe 3", should be controlled between 0.01 and 1.00 toavoid catastrophic melter failure, amuch narrower bound (e.g., between 0.01 and 0.5) is used as processinglimits. Applicability ofthese limits and their relationship to HLW composition and processing characteristics
9-10
are discussed in the following sections for WVDP and DWPF. These redox control strategies weredeveloped based on extensive laboratory, pilot, and full-scale studies.
9.3.1 West Valley Demonstration Project Redox Control Strategy
The developmentofaredoxcontrol strategyattheWVDP was divided into two parts. Firsttheredoxcouples thatinfluencethe redoxresponseinagivenglass compositionwere determined and second, the HLWfeed variables that influence the redox response of the redox couple were measured.
As discussed in the previous section, Schreiber et al. (1993) systematically studied the redox responsein simulated WVDP glasses for the most dominant redox couples and determined that the redox response inthe WVDP HLW was bounded by Fe-redox couple. Multivalent oxides present in the WVDP HLW includeFe (12 percent Fe203), Ni (0.25 wt% NiO), Mn (1 wt% MnO2), Ce (0.16 wt% CeO2 ), Cr (0.14 wt % Cr203),and Cu (0.06 wt% CuO).
At WVDP, the HLW transferred to the vitrification cell was a homogeneous mixture of washedPUREX sludge, thorium extraction (THOREX) process waste, and Cs-137-loaded zeolite (Jain and Barnes,1997). Even in this homogeneous HLW, forecasting the redox behavior based on the SLW feedcharacteristics was complex. Chemical reactions that contribute significantly to the redox response in themelter include decomposition of nitrates, nitrites, hydroxides, sulfates, phosphates, and sugar, and theevaporation of water. The concentration of sulfate, which could interfere inthe redox response, was reducedto a concentration below its solubility limit in the HLW.
Since the beginning ofnonradioactive operations in 1984, significant resources were devoted to thedevelopment of a redox model to forecast redox behavior in the melter. Parallel studies using controlledlaboratory tests, pilot-scale melter runs, and full-scale melter runs were conducted. Jain (1993) conductedlaboratory-scale studies on simulated HLW feeds to evaluate the effect of oxidizing and reducing agents onredox behavior. A laboratory-scale direct slurryredoxtest was developed to studythe redox behavior. In thistest, a known amount of liquid slurry in a covered quartz crucible is placed in a furnace at 1,150 °C for 1 hr.After 1 hr, the crucible is removed, and the glass sample is analyzed for Fe2+/Fe3". Figure 9-6 shows theeffect ofadding sugar on the redox ratio. As expected, the addition of sugar increased the Fe2+/Fe3 ratio, butFe2+/Fe3+ stabilized at 1.0. No fiurther increase in Fe2+/Fe3" was observed with additional increases in sugar.Visual examination of the sample revealed the formation of metal nuggets in the glass. This observationestablished an upper limit for sugar in the feed. In addition, the study examined the effect of dilution (additionof water) onthe redox behavior. As water was added to the slurry, more oxidizing conditions were observedinthe system. Data are shown in figure 9-7. This studyprovided information for conducting a series of well-defined tests on pilot-scale and full-scale melters.
Pilot-scale and full-scale melter runs, using simulated feed composition based on the preliminaryestimates ofthe HLW chemical composition, were conducted, and an empirical relationship between oxidizingand reducing components was developed. This relationship was expressed as
IFO = N0 3(1 - TS) (9-9)TOC
9-11
1.2
0.8 _+
a)UL
a)U-
x
MetalPrecipitation
/ | x NO 3=13.2
v I I1
0.4 _
0'-2.2 3.1 4 4.9
(gm TOC/100 gm) SlurryFigure 9-6. Effect of excess total organic carbon on the Fe'+/Fe3 + ratio at a fixed N03 content (Jain,1993)
0.21-M
C.,
LL
0.15 _
0.1 L45
0
I I I I I I
47 49 51 53 55 57 59gm H20/1 00 gm Slurry
Figure 9-7. Effect of H20 on the Fe2+/Fe3 + at a fixed total organic carbon/NO3 ratio (Jain, 1993)
9-12
where Index of Feed Oxidation (IFO) is a measure of feed oxidation, NO3 is the total nitrate concentration,TS is the fraction of total solids in the feed, and total organic carbon (TOC) is the total organic carbonconcentration. The terms TOC and N0 3(1 - TS) represent the net reduction and oxidation potential of thefeed, respectively. Note that nitrites were not included in these studies as variables because a significantquantity of nitrites was not expected. Figure 9-8 shows the Fe2"/Fe 3 ' ratio as a function of IFO. Data shownin the figure were collected using scaled vitrification system (SVS)-I and slurry-fed ceramic melter runs(Bowan, 1990). This relationship allowed the amount of sucrose needed to attain a target IFO for knownconcentrations of NO3 and TS in the feed to be estimated. For a given simulated WVDP HLW, thisforecasting method defined the IFO operating range to be between 1.6 and 2.3, which corresponds to anFe2+/Fe3+ ratio between 0.1 and 0.5.
Prior to beginning radioactive operations in July 1996, additional pilot-scale tests were conducted usingSVS-1l. The purpose was to reexamine the redox behavior based on new information available on the HLWcomposition. The new composition contained a significant amount of nitrites inthe feed, reflecting the decisionmadeto add sodiumnitrite as atank waste corrosion inhibitor. Figure 9-9 shows the revised redox forecastingcurve superimposed on the original curve obtained using feeds without nitrites in the full-scale melter. Toaccount for the nitrites in the feed, total nitrites were converted on a molar basis to an equivalent amount ofnitrates and added to the total nitrates. The results indicate that the redox state response to IFO changes wasmore rapid in the presence of nitrites inthe HLW. To avoid this rapid redox response in the melter, a decisionwas made to process the HLW at IFO > 2.7. Given the uncertainties in the chemical analysis, the target IFOof 3.0 was used to estimate the amount of sucrose required to process the feed and ensure that the Fe2̀ /Fe3 1
ratio in the melter remained between 0.01 and 0.20. The upper limit of Fe2"/Fe 3" of 0.5 was reduced to 0.2to account for the effect of nitrites, which cause a sharp increase in Fe2+/Fe3
' ratio beyond 0.2.
9.3.2 Defense Waste Processing Facility Redox Control Strategy
Similar to WVDP, the redox strategy at the DWPF evolved as the process flowsheet developed andmatured. DWPF HLW contains a high concentration of transition metals such as Fe, Ni, and Mn. Prior toadoption ofthe precipitate hydrolysis process, precipitation ofNi3S2 established the upper operating boundfor redox control (Ramsey et al., 1991). The HLW slurry consists of an insoluble portion of the waste,borosilicate glass frit, and precipitate hydrolysis aqueous (PHA) product. PHA contains the radionuclides fromthe soluble salt waste. Before it is introduced into the melter, the feed has approximately 45 percent solids.The feed contains formate, nitrate, and phenol anions that influence the redox behavior in the melter. Anextensive study with laboratory and pilot-scale testing using the scale glass melter (SGM) and integratedDWPF melter system (IDMS) was conducted to model the redox response of the feed and establish theredox relationship between laboratory tests and melter glass. Figure 9-10 shows the correlation betweenmelter feed composition and glass redox. The molar formate minus nitrate (F-N) content is plotted againstFe2"/Fe(,O,,1). Also plotted are the data collected from SGM and IDMS runs. Results indicate that redox canbe accurately predicted fromthe F-N value. Note that the redox data plotted here are shown as Fe2+/Fe(,O.,).This nomenclature has been adopted by the DWPF for representing the redox state, and no attempt wasmade to change the data for consistency purposes between WVDP and DWPF. Figure 9-11 shows thecrystalline phases formed inthe glass samples as functions ofredox and feed composition. For F-N less than1.2, no Ni3S2 is observed in the glass.
9-13
U-
NXa) .a' -0.8CY)0
Index of Feed Oxidation
Figure 9-8. The dependence of log Fe2"/Fe 3" on the index of feed oxidation: a comparison betweenslurry fed ceramic melter and pilot-scale melters (Bowan, 1990)
9-14
1
aV)
U)
C\J
U-
C)0
2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3
IFO [NO3 (1 -TS)/TOCI
Figure 9-9. West Valley Demonstration Project redox forecasting model revised after addition ofnitrites (Barnes and Jain, 1996)
9-15
0.8
0.71
0.61
ai)U-~I
Na)
LL
0.51
0.41
o - Crucible Tests]
00
0
0
00
00
0
0
|PHA2| 0 O
H G1 0 ISGM 61
BL 2 F _ I21
IBL 1 III
0.31
0.21
0.1
('hr%
0.0 0.5 1.0 1.5 2.0Formate - Nitrate (Mol)
2.5 3.0
Figure 9-10. Comparison of scale glass melter and integrated Defense Waste Processing Facilitymelter system campaigns with crucible redox data (Ramsey et al., 1991)
0.8
Ni3S2, Cu2S, Ni-Cu0.7 _
0.6k NiFe 2O4 and/or Ni3S2
00
0
00 0
0.51-
0
0
0a)
L-
('Ja)IL
A lr
0.3 _
0.21_NiFe2 04 Spinel
0
0
0 0
0
00.1 F
0.0 9 I0.5 1.0 1.5 2.0 2.5 3.0
Formate - Nitrate (Mol)
Figure 9-11. Comparison of crystalline phases with slurry (F-N) and glass redox (Ramsey et al.,1991)
9-16
As the process matured, DWPF determinedthatthe expected radiation doses fromrthe Cs-137 in theprecipitate required additional copper formate catalystthaninitiallyanticipated for complete PHA formation(Schumacker and Ramsey, 1994). This increase in copper concentration in the melter feed increases thepotential for precipitation of metallic copper phases in the melter, which could accumulate over a period oftime and interfere with the joule heating. Anticipated levels of elemental copper in glass are 0.4 to 0.5 wt%.Figure 9-12 shows a plot of Cu concentration versus F-N. Open circles indicate glasses with no precipitatedCu, while solid circles indicate glasses with precipitated Cu. Boundary lines shown in the figure representsthe limits for maintaining Cu as a soluble species in glass. Based on this work, the following limits wereestablished:
* If Cu • 0.45 wt%, F-N should not exceed 1.4* If Cu is between 0.45 and 0.70 wt%, F-N should be s 0.5* Cu should not exceed 0.7 wt%
Because of Cu in the HLW, DWPF is forced to operate the melter in a fairly oxidizing range(Fe 2+/Fe(,.w) •0.03). Almost continuous formation of foam in the melter has resulted.
9.4 REDOX ISSUES FOR TANK WASTE REMEDIATION SYSTEM
The flowsheet proposed by BNFL Inc. uses sugar as a reducing agent for controlling redox in theLAW and HLW melters. Separate redox control strategies are needed for LAW and HLW feeds becausethe characteristics of the two feed streams are quite different. The LAW stream is stripped of almost allmultivalent elements and radioactive species, but Fe is added as a glass former. The HLW stream hassignificant concentration of multivalent elements such as Fe, Mn, Ni, and Cr and has most ofthe radioactivespecies. Additional complexity in developing a redox strategy is imposed by the variability in the feedcomposition at Hanford. Four waste streams of different compositions will be processed and separated intoHLW and LAW streams, requiring a different redox control strategy for each waste stream. If waste streamsare homogeneously mixed prior to feeding to the melter, one redox strategy could be sufficient. As indicatedin the previous section, development of a redox control strategy is a complex process. Attention needs to befocused on all potential reducing and oxidizing agents as well as multivalent elements irrespective of theirconcentrations. Lessons learned from the development of the redox control strategy at WVDP and DWPFcould be beneficial in developing a suitable redox strategyfor Hanford wastes. For example, minor additionsof material such as nitrite in WVDP HLW or Cu in DWPF HLW could severely restrict the redox operatingrange.
Recent advances in the development of in-situ redox measurement techniques could furtherfacilitate redox measurement in the melter. For example, a redox probe developed by Kuhnreich & Meixner(Korwin and Pye,1995) was tested at WVDP in an SVS-III and used by several commercial glassmanufacturers (Muller-Simon and Mergler, 1988). Theutilityofsuchprobes is limited becausetheycannotforecast redox and determine the amount of reductant needed to attain target redox in the melter. The redoxprobe, however, is a useful tool to provide a check on melter redox condition and to allow a prompt shutdownof a melter in case the redox response exceeds the operating range. Thus, the probe should be considereda monitoring sensor.
9-17
1.2
1.1
0
Ca)Ea)
wU)enCZ,
Ua)Ca
0C.)
I-
a)
00
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.21 to-0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5
Molar (Formate - Nitrate) in Melter Feed
Figure 9-12. Effect of feed chemistry and copper content on the likelihood of formation of insolublecopper precipitates during vitrification (Ramsey et al., 1994)
9-18
9.5 SUMMARY
The redox state in a melter is probably the most important process control parameter. A redoxresponse within a specified range is necessary to avoid process problems that eventually could lead topermanent melter damage. The operating range for redox in a melter depends on the amount and type ofmultivalent elements present in the HLW. Because Fe is abundant in most wastes and oxidation states canbe easily measured, the redox response in waste vitrification systems is determined by the Fe2"/Fe 3"ratio. However, the influence of some minor components inthe waste, such as Cu and Mn, cannot be ignored.A glass melt is defined as extremely reducing if the Fe2 1/Fe3+ ratio is greater than 1. Under such conditions,sufficient conductive metals and metal sulfides could accumulate and short-circuit the melter. If the Fe2"/Fe 3"ratio is less than 0.01, a glass melt is defined as extremely oxidizing. Under extremely oxidizing conditions,foaming is observed in the melter. Foam creates an insulating layer of gas bubbles between the cold cap andthe melt, disrupting the thermal gradients in the melter. If controlled at its onset, foaming typically slows theglass production rate and is not an issue. If foaming is not controlled, however, conditions can arise that couldlead to melter shutdown. Eventhough Fe precipitation is not observed until the melt attains an oxygen fugacityof 10-9 (Fe2+/Fe 3+ ratio of 1), the upper limit for acceptable redox range is limited to an oxygen fugacity of10-7. This value corresponds to an Fe2+/Fe3+ ratio of 0.5. This lowering of the upper limit for an acceptableredox range is attributed to the presence of sulfur in the HLW slurries, which establishes a S6+/S2 redoxcouple in the system and results in the formation of electrically conductive sulfide phases such as FeS andNiS. These phases could electrically short-circuit the melter. If the oxygen fugacity is less than 10 02, redoxcouples such as Ce, Mn, and Cr begin to influence the redox response in the system causing formation offoam inthe glass melt. Therefore, the redox response for processingnuclear waste glasses has been boundedbetween the oxygen fugacity of I 0-4 and 10-7, which corresponds to Fe2"Fe 3" ratios of 0.01 and 0.5. Redoxequilibrium in glass melts depends on the oxygen ion activity in the melt, the equilibrium constant [K(T)],which is a function of the Gibbs free energy and temperature, the redox couple activity; and the oxygenactivity [O.
Thepresence of multiple redox couples, along with anumber of reducing and oxidizing agents intheHLW, makes redox forecasting and control in a HLW melter complex and challenging. Because the HLWfeeds typically contain significant concentrations of oxidizing agents such as nitrates and nitrites, the selectionof a proper reducing agent is important. The selection ofthe type and amount of reducing agent depends notonly onits decompositiontemperature, but also onthe decomposition temperature of oxidizing agents. Theeffectiveness of the reducing agent could be compromised if the decomposition temperature differential islarge. Starch, sugar, formates, urea, or carbon black could react differently in the same feed.
Both WVDP and DWPF have invested significant resources in understanding and developing redoxforecasting and control strategy, and the lessons learned could be beneficial in developing a suitable redoxstrategy for Hanford wastes. For example, minor additions such as nitrite in WVDP HLW or Cu in DWPFHLW could severely restrict the redox operating range
9.6 REFERENCES
Barnes, S.M., and V. Jain. Vitrification systems testing to support radioactive glass production at the WestValley Demonstration Project. Proceeding of the WasteManagement '96 Symposia. Tucson, AZ:Waste Management Symposia Inc. 1996.
9-19
Bickford, D.F., and R.B. Diemer, Jr. Redox control of electric melters with complex feed compositions.Journal of Non-Crystalline Solids 84: 276-284. 1986.
Bowan, B.W. Redox forecasting correlation developed using a new one-tenth area scale test melterfor vitrifying simulated West Valley high-level radioactive wastes. Master's thesis, AlfredUniversity, Alfred, NY. 1990.
Jain, V. Redox forecasting in the West Valley vitrification system. Proceedings of the Third InternationalConference on Advances in Fusion and Processing of Glass, New Orleans, Louisiana,June 10-12, 1992. Ceramic Transactions Volume 29. A.K. Varshneya, D.F. Bickford, and P.P.Bihuniak, eds. Westerville, OH: American Ceramic Society: 523-533. 1993.
Jain, V., and S.M. Barnes. Radioactive waste glass production at the West Valley Demonstration Project.Proceedings of the WasteManagement '97Symposia. Tucson, AZ: Waste Management SymposiaInc. 1997.
Korwin, D.M., and L.D. Pye. Milestone #3 Final Report on the Oxygen Activity Measurement ofReference 6 Glass. Alfred, NY: Alfred University. 1995.
Muller-Simon, H., and K.W. Mergler. Electrochemical measurements of oxygen activity of glass melts inglass melting furnaces. Glastech Ber 61(10): 293-299. 1988.
Paul, A., and R.W. Douglas. Ferrous-ferric equilibrium in binary alkali silicate glasses. Physics andChemistry of Glasses 6(6): 207-212. 1965.
Ramsey, W.G., C.M. Jantzen, and D.F. Bickford. Redox analyses of SRS melter feed slurry: Interactionsbetween nitrate, formate and phenol based dopants. Proceedings of the Fifth InternationalSymposium on Nuclear Waste Management IV 9 3 rd Annual Meeting of the American CeramicSociety, Cincinnati, Ohio, April 29-May 3, 1991. Ceramic Transactions Volume 23. G.C. Wicks,D.F. Bickford, and L.R. Bunnell, eds. Westerville, OH: American Ceramic Society: 259-265. 1991.
Ramsey, W.G., N.M. Askew, and R.F. Schumacher. Prediction of Copper Precipitation in the DWPFMelter from the Melter Feed Formate and Nitrate Content. WSRC-TR-92-3 85. Revision 0.Aiken, SC: Westinghouse Savannah River Company. 1994.
Ryan, J.L. Redox Reactions and Foaming in Nuclear Waste Glass Melting. PVTD-C94-03.02K.Richland, WA: Pacific Northwest National Laboratory. 1994.
Schreiber, H.D. Redox processes in glass-forming melts. Journal of Non-Crystalline Solids 84:129-141.1986.
Schreiber, H.D., and A.L. Hockman. Redoxchemistryin candidate glasses for nuclear waste immobilization.Journal of the American Ceramic Society 70(8): 591-594. 1987.
9-20
Schreiber, H.D., and E.V. Sampson, Jr. A corrosion model for metals in molten slags. Proceedings ofthe Corrosion of Materials by Molten Glass Symposium. 98th Annual Meeting of the AmericanCeramic Society, Indianapolis, Indiana. Ceramic Transactions Volume 78. G. Pecoraro,J.C. Marra, and J.T. Wenzel, eds. Westerville, OH: American Ceramic Society: 3-15. 1996.
Schreiber, H.D., S.J. Kozak, R.C. Merkel, G.B. Balazs, and P.W. Jones, Jr. Redox equilibria and kinetics ofiron in a borosilicate melt. Journal of Non-Crystalline Solids 84: 186-195. 1986a.
Schreiber, H.D., S.J. Kozak, A.L. Fritchman, D. S. Goldman, and H.A. Schaeffer. Redoxkinetics and oxygendiffusion in a borosilicate melt. Physics and Chemistry of Glasses 27(4): 152-177. 1986b.
Schreiber, H.D., C.W. Schreiber, M.W. Riethmiller, and J.S. Downey. The effect of temperature of theredox constraints for the processing of high-level nuclear waste into a glass waste form.Symposium Proceedings for the Materials Research Society. Symposium Proceedings 176.Pittsburgh, PA: Materials Research Society: 87-94. 1990.
Schreiber, HD., C.W. Schreiber, and C.C. Ward. Redoxsystematics in model glass compositions from WestValley. Proceedings for the Materials Research Society Symposium. SymposiumProceedings 294. Pittsburgh, PA: Materials Research Society: 87-94. 1993.
Schumaker, RF. and W.G. Ramsey. Conditions for precipitation of copper phases in DWPF glass.Proceedings of the Symposium on Environmental and Waste Management Issues in the CeramicIndustry, presented at the 9 5th Annual Meeting of the American Ceramic Society, Cincinnati,Ohio, April 18-22,1993. Ceramic Transactions Volume 39. G.B. Mellinger, ed. Westerville, OH:American Ceramic Society: 249-256. 1994.
9-21
10 NOBLE METALS IN HIGH-LEVEL WASTE GLASS MELTS
Noble metals such as Ru, Rd, and Pd intheHLW originate from the fission of U-235. Table 10-1 shows theradioactive noble metal isotopes that are present in spent fuel. Ru is the most abundant noble metal in theHanford HLW.
Noble metals in the borosilicate glass melts have been a major concern in the HLW vitrification processbecause of their low solubility, high volatilization rate, and high electrical conductivity. In 1985, theaccumulation of noble metals on the floor of the Pamela melter, Mol, Belgium, resulted in electricalshort-circuiting of ajoule-heated melter. Elliottet al. (1994) reviewed various methods for separating noblemetals from the HLW. However, efforts to apply noble metal separation methods for Hanford HLW wereabandoned because of the complex composition ofthe HLW that interfered with the separation process. Inthis chapter, a discussion on behavior of noble metals in glass melts and the melter is followed by a reviewofvarious studies on the subject and the implication ofnoble metals onHLWvitrification at the Hanford site.
10.1 NOBLE METAL SOLUBILITY IN GLASS MELTS
Borosilicate glass-based waste formhas averylimited noble metal solubility. In silicate glasses, thesolubilities ofRh, Pd, and Ru are about 0.05,0.03, and 0.01 wt%, respectively. Noble metal concentrationsgreater than solubility limits produce metal particles suspended in the melt. These metals are either flushedout with the glass melt during pouring or settle on the floor of the melter. Noble metal settling on the melterfloor depends on factors such as
* Residence time in the melter. The longer the residence time in the melter, the higher theconcentration of the noble metals that settle on the floor of the melter.
* Convective currents in the melter. Stronger convective currents keep the noble metals suspendedinthe melt. Melter idling is the worst-case scenario. During melter idling, convective currents areweaker, which allow noble metals to slowly settle on the melter floor. The settling rate dependson the particle size of noble metals. The larger the particle size, the quicker the particles reachthe melter floor.
Table 10-1. List of noble metal isotopes present in high-level waste (Elliott et al., 1994)
I ~Isotope I Half-Life | Stable Decay Product
Ru-103 39.3 d Rh-103
Ru-106 372.6 d Pd-106
Rh-102 2.9 yr Ru-102
Rh-103m 56.1 nun. Rh-103
Pd-107 6.5 E+6 yr Ag-107
10-1
In addition, both residence time and magnitude of convective currents depend on the viscosity ofglass. Noble metal concentrations in various HLW glasses for NCAW, which is now referred to as EnvelopeB/D waste, DWPF, Kernforschungszentrum Karlsruhe (KfK), Germany, and WVDP are shown intable 10-2.
10.2 NOBLE METAL BEHAVIOR IN THE MELTER
Since the failure ofthe Pamelamelter, Mol, Belgium due to noble metal accumulation onthe melterfloor, several studies have been performed at various plants and research laboratories. These studies arebriefly summarized in this section.
KfK examined noble metal retention in the melter using various melter designs. Their studies aresummarizedinareportbyElliott et al. (1994). The studies indicated thatthe accumulation ofnoble metals canbe significantly reduced by increasing the slope ofthe melter walls. For a flat-bottom melter with a bottomdrain, approximately 65 percent of the noble metals were retained in the melter. However, with 750/600sloped melter walls, electrical resistance did not indicate any accumulation of noble metals. KfK alsocompared the effect of a pour system on noble metal retention. The bottom drain retained only 10.3 percentRu, while the overflow system retained 41.4 percent Ru. An overflow system is currentlyused in the WVDPand DWPF melter designs for glass pouring. The studies also found that in a melter with 450 sloped walls,convective currents in the glass melt decreased the Ru retention from 38 to 24 percent. In a 75 °/60° slopedwall, no measurable effect was observed.
During nonradioactive operations at the WVDP between July 1986 and May 1988, approximately9.3 kg of RuO2 was fed to the melter. The analysis of the samples retrieved from the melter floor indicatedan almost 6-cm deep sludge layer on the melter floor. Approximately 87 percent ofthe RuO2 fed to the melterwas flushed from the melter during normal glass pouring. The WVDP melter has sloped walls (Jain et al.,1991). The sludge layer consisted of mostly Fe-Cr spinel and RuO2 crystals dispersed in the matrix. Inaddition, crystalline Ce203 and undissolved phases consisting of A1203 and Cr203 were inhomogeneouslydistributed and were presentintrace amounts. Fe-Cr spinel represented almost33 vol % ofthe sludge. Basedon these results and the total inventory of noble metals in the HLW at WVDP, it was determined that the
Table 10-2. Noble metal concentration in various high-level waste glasses (Elliott et al., 1994)
Defense West ValleyNeutralized Waste Demonstration
Noble Current Acid Processing Kernforschungszentrum ProjectMetal Waste Facility Karlsruhe (Jain et al., 1991)*
RuO 2 0.11 0.1 0.655 0.10
Rh 0.024 0.02 0 0.03 (as oxide)
Pd 0.03 0.03 0.298 0.05 (as oxide)
*Jain, V., S.M. Barnes, T.K Vethanayagam, and L.D. Pye. Noble metal and spinel deposition on the floor of the joule-heated
melter. Journal ofthe American Ceramic Society 74(7): 1,559-1,562. 1991.
10-2
entire batch ofthe HLW containing noble metals can be processed using current melter design (with slopedwalls and overflow system for glass pouring) without impactingthe electrical resistance ofthe melter. Duringradioactive operations, however, electrical resistance between the side and bottom electrodes dropped. Thisdecrease is being attributed to preferential accumulation of noble metals on the melter floor but no reportshave been published on the subject.
In 1990, PNNL conducted gradient furnace testing to estimate the behavior of noble metals in a coldcap (Anderson et al., 1994). Dried feed in a long-sample boat was heated in a 590 to 940 'C gradient. Theresults indicated some agglomeration ofnoble metals at the cold end (590 'C) ofthe sample boat, while 1 -gmRuO2 particles with only a few agglomerates were observed at the hot end (940 °C). This study was followedbytesting noble metal behavior in the / 100th scale research scale melter (RSM) in 1992 (Elliott et al., 1994).The RSM operated for 48 d during which time the effects of various operating parameters such as glass andplenum temperature, oxide concentration, redox, and noble metal concentrations were evaluated. A reductionin electrical resistance was observed duringthe latterpart ofthe campaign. Melter evaluation after operationsindicated the presence of a 2-4-mm-thick (0.08-0.16 in.) noble metal layer on the melter floor, which causedone-third of one paddle electrode to corrode or dissolve. The electrode dissolution was attributed to localizedheating due to short-circuiting of the electrode to the noble metal layer. In addition, noble metal causeddissolution of 1.3 cm (0.5 in.). of refractorylayer on the floor. Mass balance results indicated that 5 percentof the noble metals were retained during segments where noble metal concentration was nominal and46 percent of the noble metals were retained during the segments with double the nominal concentration ofnoble metals. Such a large accumulation and its impact on melter components were surprising because theresidence time in the RSM is 10 percent ofthe full-scale melter. The analysis of glass samples poured fromthe melter indicated 10-grm average diameter needle-like RuO2 particles with 100-200-pm RuO2agglomerates. RuO2 agglomerates increased with an increase in the noble metal concentration in the feed.
The RSM study was followed by an engineering-scale melter (ESM) study. The ESM melter wasone-tenth scale and was operated by KfK in Germany. The ESM operated for 49 d with noble metals. Themass balance results indicated retention of 35 percent of the noble metals on the melter floor. In addition,during the last segment of the campaign, the electrical resistance between the lower set of electrodesdecreased by 10 to 15 percent. The glass poured from the melter indicated similar crystals in the glass, butthe average diameter ofthe particles was slightly larger than RSM samples. PNNL also conducted computermodeling using the Transient Energy, Momentum, Pressure Equations Solutions in Three Dimensions(TEMPEST) Code. Results are discussed by Elliott et al. (1994).
Table 10-3. Amount of noble metals processed through integrated Defense Waste ProcessingFacility melter system (Hutson, 1994)
I Noble Metal I Amount (kg)
Ru 15.68
Pd 5.73
Rh 3.12
10-3
The DWPF scale melter system, IDMS was used to study noble metal behavior in the SavannahRiver Site (SRS) and Hanford wastes. Three types of SRS wastes (Blend, HM, and PUREX) and one typeof Hanford waste (NCAW) were processed between June 1990 and March 1993. The total amount of noblemetals processed through the melter is shown in table 10-3. Pd and Rh were added as nitrate solutions, whileRu was added as nitrosylruthenium hydroxide [RuNO(OH)3]. Samples taken from the melter floor duringoperations indicated increasing amounts of noble metals onthe melter floor withtime. In addition, the samplesretrieved from the center and from the edges ofthe melter indicated nonuniform distribution of noble metalson the floor. The samples from the center contained Ru and Rh oxides and large amounts of spinel while thesamples from the outer edge contained fewer spinel, metallic Pd and Pd tellurides, Ru and Rh oxides, andRuS2. After approximately 550 hr of operation, the ratio of upper melt resistance and lower melt resistanceincreased. The melter floor sample indicated a good correlation between the increase in resistivity ratio andaccumulation of noble metals on the floor. The overall mass balance indicated 35-, 21-, and 0-percentretention of Ru, Rh, and Pd in the melter, respectively (Elliott et al., 1994).
Studies indicate that a melter needs to be designed to accommodate anticipated accumulation of noblemetals onthe melter floor withoutaffectingthe current distribution inthe melter. The use of sloped walls andpositioning electrodes away from the floor could allow significant amounts ofnoble metals to accumulate onthe floor before a resistance drop is observed in the melter. Further, melter idling, which causes suspendednoble metals to settle on the floor, should be minimized.
10.3 TANK WASTE REMEDIATION SYSTEM CONCERNS
The settling of noble metals on the floor of the melter could cause premature failure of the melter.At Hanford, HLW will contain most of the noble metals in the feed. The HLW melter should incorporatelessons learned from the failure of the Pamela melter, operation ofthe radioactive WVDP melter, and otherresearch and pilot-scale melters. Lessons learned include
* Using of sloped walls in the melter* Positioning electrodes away from the floor* Minimizing melter idling time* Regular monitoring of electrode current
10.4 SUMMARY
Noblemetals such as Ru, Rd, and PdintheHLWoriginatefromthefissionofU-235. Ruisthemostabundant of all the noble metals in the Hanford HLW. During glass melting, these metals have been a majorconcern because of their low solubility, high volatilization rate, and high electrical conductivity. In 1985, theaccumulation of the noble metals on the floor of the Pamela melter in Mol, Belgium, resulted in electricalshort-circuiting of ajoule-heated melter. Borosilicate glass-based waste form has averylimited noble metalsolubility. In silicate glasses, solubility of Rh, Pd, and Ru are about 0. 05, 0.03, and 0.01 wt%, respectively.Noble metal concentration greater than solubilitylimits results in metal particles suspended inthe melt. Theseparticles are either flushed out with the glass melt during pouring or then settle on the melter floor. Noblemetal settling on the melter floor depends on factors such as residence time in the melter, convective currentsin the melter, and melt viscosity. Melter idling is the worst-case scenario for noble metal accumulation. Duringmelter idling, convective currents are weaker, which allows noble metals to settle slowly onthe melter floor.
10-4
The settling rate depends on the particle size of noble metals. The larger the particle size, the quicker theparticles reach the melter floor. Noble metals behavior evaluated at WVDP, KfK, Pamela, DWPF, andPNNL indicates that a melter should be designed to accommodate the accumulation of noble metals on themelter floor without affecting the current distribution in the melter. The use of sloped walls and positioningelectrodes away from the floor could allow significant amounts of noble metals to accumulate on the floorbefore a resistance drop is observed inthe melter. Further, noble metal settling can be minimized by reducingthe idling time in the melter.
10.5 REFERENCES
Anderson, L.D., T. Dennis, M.L. Elliott, and P. Hrma. Noble metal behavior during melting of simulatedhigh-level nuclear waste glass feeds. Proceedings of the Symposium on Environmental and WasteManagement Issues in the Ceramic Industry. 95 th Annual Meeting of the American CeramicSociety, Cincinnati, Ohio, April 18-22, 1993. Ceramic Transactions Volume 39. G.B. Mellinger,ed. Westerville, OH: American Ceramic Society: 265-272. 1994.
Elliott, M.L., L.L. Eyler, L.A. Mahoney, M.F. Cooper, L.D. Whitney, and P.J. Shafer. PreliminaryMelterPerformance Assessment Report. PNL-9822. Richland, WA: Pacific Northwest NationalLaboratory. 1994.
Hutson, N.D. The behavior oftheplatinumgroup metals in aborosilicate waste glass andtheir effects ontheoperation ofajoule-heated ceramic melter. Proceedings of the Symposium on Environmental andWaste Management Issues in the Ceramic Industry. 9 5 th Annual Meeting of the AmericanCeramic Society, Cincinnati, Ohio, April 18-22, 1993. Ceramic Transactions Volume 39.G.B. Mellinger, ed. Westerville, OH: American Ceramic Society: 257-254. 1994.
Jain, V., S.M. Barnes, T.K. Vethanayagam, and L.D. Pye. Noble metal and spinel deposition onthe floor ofthe joule-heated melter. Journal of the American Ceramic Society 74(7): 1,559-1,562. 1991.
10-5
11 SULFUR SOLUBILITY IN WASTE GLASSES
Borosilicate glass-based waste forms have limited sulfur solubility. In the WVDP and DWPF waste forms,S03 levels are maintained below 0.25 wt% to avoid formation of an immiscible sodium sulfate phase in themelt. Traditionally, wastes containinghigh concentrations of sulfur are washed with water to remove solublesulfate salts. The salt solution is then treated separately as a low-level waste. The HLW at WVDP and atthe Pamela vitrification facility, Mol, Belgium, were washed several times with water to reduce S03
concentration below 0.25wt%. If the concentration of S03 is in excess of 0.25 wt%, the SO3 concentrationlimits the waste loading in the glass. Any improvement in increasing the waste loading through betterunderstanding of sulfur solubility could result in significant cost savings. Current plans at Hanford involvedevelopment of glass compositions for LAW that can accommodate SO3 concentration in excess of 1 wt%.Expected concentration of SO4 from various feed streams at Hanford is shown in table 1 1-1. In this chapter,a brief discussion on the sulfur solubilitylimit imposed on the glass composition is followed by a review ofvarious sulfur solubility studies and sulfur solubility implication on the LAW vitrification at the Hanford site.
11.1 SULFUR SOLUBILITY LIMITS IN HIGH-LEVEL WASTE BOROSILICATEGLASSES
Borosilicate glass-based waste forms have limited sulfur solubility. Eventhough SO3 solubilityunderoxidized conditions is approximately I wt%, the SO3 levels inthe WVDP and DWPF glasses are maintainedbelow 0.25 wt%to avoid formation of an immiscible sodium sulfate phase inthe melt because SO3 solubilityin glass is a strong function ofmelt redox conditions. SO3 solubility decreases sharply as glass melts becomereducing. The SO3 solubility limit is established based on the operating redox range of 0.01 and 0.5.
The sodium sulfate phase, usuallyreferred to as "gall," is volatile, water soluble, and floats on the meltsurface. Formation of gall could result in undesirable consequences such as
* Partition of radionuclides into the gall phase* Foam formation in the melter* Excessive corrosion of refractories and Alloy 690 components in the melter* Disruption of heat balance in the melter (gall acts as an insulating layer on the surface of the
melt)
Table 11-1. SO4 concentration limits for A, B, and C low-activity waste feed envelopes (Pabalanet al., 1999)
I ~ Envelope SO 4 mol/mol of Sodium
A 9.7 x 10-3
B 7.0 x 10-2
C 9.7 x10-3
11-1
11.2 SULFUR SOLUBILITY STUDIES
11.2.1 Effect of Redox
Schreiber et al. (1994) studied the effects of redox on S03 solubility in WVDP and DWPF HLWglasses. Figure 11-1 shows the effect of oxygen fugacityon S0 3 solubility. The relationship between oxygenfugacity and Fe2+/Fe34 ratio in glass is discussed in chapter 9. The V-shaped behavior indicates two differentmechanisms by which sulfur is incorporated into the glass melt. Minimum S03 solubility occurs at an oxygenfugacity of 1 o-8 8. At an oxygen fugacity higher than 10 -8.8, sulfur is incorporated in the glass as sulfate ion.
At an oxygen fugacity lower than 10 -8 8, sulfur is incorporated in the glass as sulfide ion. In glassmelts, as oxygen fugacity falls below 10-11, a S6+/S2- redox couple in the system results in the formation ofenough electrically conductive sulfide phases such as CuS, FeS and NiS electrically short the melter. At anoxygen fugacity above 10-2, redox couples such as Ce, Mn, and Cr can influence the redox response in thesystem and cause formation of foarn in glass melts.
Sulfur Solubility in Glasses
Sullivan (1995) studied the retention of sulfur in a Hanford HLW glass as a function of sulfurconcentration, basicity, A12 03 , and P2 05 concentration. Figure 11-2 shows the distribution of S03 in solublegall, glass and volatilized fraction as a function of S0 3 added to the batch. Results indicated 1 00-percent
wt% S0.001 0.01 0.1 1 10
16 lI I l I I I I I
14 -
12 -
10C\j
0 Target6 - |SO2- <??? Processing4 < ? - Range
2 -
-3 -2 -1 0 +1log (wt% S)
Figure 11-1. Sulfur redox/solubility systematics in WVF-17993 (solid line), in SRL-131 (dottedline) and in SRL-131 + 10 wt% Fe (dashed line) as a function of the imposed oxygen fugacity in anatmosphere of 15 vol% SO2 at 1,150 °C (Schreiber et al., 1994)
11-2
100
* 75 glass
0
n50
n loss
0 25
filtrate
0.5 1 1.5 2As Batched SO3 (wt%)
Figure 11-2. S0 3 distribution in soluble gall, glass, and volatilized fraction as a function ofS03 content (Sullivan, 1995)
retention of sulfur in glasses with S03 concentration less than 1.0 wt%. The glass compositions studied hadlow (1-2 wt%) Fe203 concentration and were melted in oxidizing conditions (redox was not controlled inthemelt). For S03 concentrations greaterthan 1.3, ayellowlayer of gall was observed, and 1.1 to 1.2 wt% S03was retained in the glass. In a base glass containing 1.6 wt% S0 3, S03 retention decreased with an increasein A1203 concentration (figure 11-3) and increased with an increase in P205 concentration (figure 1-4). Inall cases, gall was observed on the surface.
Li et al. (1 995) studied the kinetics of gall formation using plutonium finishing facility sludge and WestValley glass composition WV-1 83 spiked with up to 2 wt% sodium sulfate. The glass compositions wereclassified as high Na2O and low B203, and low Na2O and high B203. In high Na2O and low B203 glasses,phase segregation occurred before melting was complete. In glasses containing low Na2O and high B203, nophase segregation was observed during melting. InWV-1 83 glass, phase segregation did not occur in batchescontaining s 1 wt% S03, while a yellowish phase formed at higher concentrations. The results indicated thatphase segregation is a kinetic process occurring inthe early stages of batch melting. The glass containing highconcentrations of boric acid suppresses sulfate segregation, while A1203 promotes sulfate segregation. Inbatches containing both sulfur and phosphorus, the sulfate phase dissolves in the phosphate phase and inducesphosphate segregation. While sulfate segregation was suppressed byhigh boric acid concentration, excessivefoaming was observed in the melts.
11-3
80
6000
-ZC0
* 40.4-
Men
oft 20Un
rf%*
glass
filtrate
'j loss
0 ~4 8 12 6Al 2 03 (wt%)
Figure 11-3. S03 distribution in soluble gall, glass, and volatilized fraction as a function ofA1203 content (Sullivan, 1994)
100I
0
0cC'COVa)-0
.C_
a)
75 .
50
glass
filtrate
loss
25
01L) 2 4
P 20 5 (wt%)6 8
Figure 11-4. SO3 distribution in soluble gall, glass, and volatilized fraction as a function of P20,content (Sullivan, 1994)
11-4
While sulfate segregation is a kinetic process, the sulfate retention is a thermodynamic process.Li et al.' analyzed sulfate retention in the melt and described the dissolution of sulfate melt as
=Si -O-Si=+2Na+:0 2- [glass]-->2(=- O:Na')[glass]
S03 [gas] + 2Na+ 02- [molten salt] +-~--+ 2Na+: S02- [molten salt]
where aSi- 0- Sia and - Si- 0 - are the BO and NBO in glass, respectively, and ki is the reaction equilibriumconstant. Based on the above reaction equilibrium, the solubility (S[S031 ) of sulfate in a silicate melt, asexpressed in terms of S03 concentration is
2 22
SIO = Ps, k2 Na+ a =NBOBQ (11-2)S 3] s kI aBO'Y1NaSO PSO3 CBO
where k, and k2 are the equilibrium constants of the above reactions, pso is the S03 partial pressure, a.
(i = Na, NBO, BO) is the il species activity, Y Na2SO is the Na2SO4 activity coefficient, and xv includes theactivity coefficients of Na, Na2SO4, NBO, and BO, the reaction equilibrium constants (k1 and k2), andconcentration of sodium ions. At agiven temperature and Pso3 ,Eq. (11-1) indicates that SS3 is a linear
fimction of the (CNB0 )2/CBo ratio.
Figure 11-5 shows measured SO3 retention concentrations in glass versus (CNB0 )2 CB0. The sulfateretention in glass increases nonlinearly as the (CNBO)2 /CBO ratio increases, and the sulfate retention linearlyincreases as (CNBO)2 CBO decreases with increasing P205 concentration. While the sulfate retention in thesecomplex systems can be correlated with (CNB0)2 ICBO, P205 has a different effect on sulfate retention in theglass. Increase in sulfate retention by adding phosphate is attributed to the formation of a separated phasein which sulfate and phosphate coexist. The figure also includes data for other borosilicate and silicatecompositions reported inthe literature. Despite differences in the melting temperature ofthese glasses, sulfateretention concentration falls on the same curve. Li et al.2 also studied sulfate segregation tendencies andshowed that even a melt with high sulfate retention could result in sulfate segregation. They attributed thesulfate segregation tendency to alkali and sulfate reactions occurring prior to the formation ofthe glass melt.
High S0 3 solubility in phosphate glasses has been observed by Stefanovsky et al. (1995). Theystudied the 44Na2O, 20A1203, (36-x)P205 , andxSO3 system for immobilizinghighsulfateradioactive wastes.The results indicated that glasses containing up to 8.5 mol% S0 3 were vitreous and showed no sulfate phasesegregation. While this system can accommodate higher sulfate concentration inthe melt, its applicabilitytomeet product qualification requirements for disposal is not known.
1Li H., P.R. Hnna, and J.D. Vienna Sulfate retention and segregation in simulated waste borosilicate glasses. CeramicTransactions of the Environmental Issues and Waste Management Technologies in the Ceramic and Nuclearlndustries VI, St. Louis,Missouri, April 30-May 3, 2000. Ceramic Transactions Volume 119. Westerville, OH: American Ceramic Society. In press.
2 Ibid.
11-5
2.0
1.8 ./4 WIN--- --------. ,............0 P series A Silicates
WL series1.6A
A
1.4 C C A
W12~~~~~~~~~~~ AL(I)~~~~~~~~I
1.2
- -E~ mo
'S 0.6 0l0)4
0.2
0.00 5 10 15 20 25 30 35
CNBO/CBO
Figure 11-5. Sulfate retention in glass versus (CNBO) 2/CBO, where CNBO and CBO are calculatedmol concentrations of nonbridging oxygen and bridging oxygen 3
Recently, Davis et al.4 developed glass compositions for high-sulfate (2-3 wt% SO3 ), high-lead
(9.0 wt% PbO) wastes stored in Silos 1 and 2 atFernald. Theyadjustedthereductant additions and visually
observed the salt formation and metal precipitation during vitrification. The data are shown in table 11-2.
The data show that by increasing the reductant addition to the batch, the salt on the surface of the
melt decreases and eventually disappears. However, metal droplets start appearing before all salt is dissolved
in the melt. These data are in contrastto the sulfate solubilitybehavior shown in figure 11-1. Itis possible that
the glass under study contains approximately 9.0 wt% PbO and no B2 03 , which could result in sulfate
solubility behavior different from the typical borosilicate glasses. The applicability of these data to the
3Li, H., P.R. Hnna, and J.D. Vienna. Sulfate retention and segregation in simulated waste borosilicate glasses. CeramicTransactions of the Environmental Issues and Waste Management Technologies in the Ceramic and NuclearIndustries V1, St. Louis,
Missouri, April 30-May 3, 2000. Ceramic Transactions Volume 119. Westerville, OH: American Ceramic Society. In press.
4Davis, D.H., D.M. Bennert, and E. Nicaise. Control of batch redox to permit practical manufacturing of Femald POPTsurrogate-based glass. Ceramic Transactions of the Environmental Issues and Waste Management Technologies in the Ceramic and
Nuclear Industries VI, St. Louis, Missouri, April 30-May 3, 2000. Ceramic Transactions Volume 119. D. Spearing and V. Jain, eds.Westerville, OH: American Ceramic Society. In press.
11-6
Table 11-2. Formation of salt layer as a function of redox in S0-1-10 glass (Davis et al., 2000)*
Wt% Equivalent C I Sulfate Salts Present? I Metal Present? I Redox Fe2+/IFe
0.0 Yes-Heavy Amount None 0.03
0.1 Yes-Heavy Amount None 0.10
0.2 Yes-Moderate Amount None 0.18
0.3 Yes-Minor Amount Yes 0.19
0.4 None Yes 0.20
*Davis, D.H., D.M. Bennett, and E. Nicaise. Control of batch redox to permit practical manufacturing of Fernald POPTsurrogate-based glass. Ceramic Transactions of the Environmental Issues and Waste Management Technologies in the Ceramic
and NuclearIndustries VI, St. Louis, Missouri, April 30-May 3, 2000. Ceramic Transactions Volume 1 9. D. Spearing and V. Jain,eds. Westerville, OH: American Ceramic Society. In press. 2000.
borosilicate borosilicate type is questionable. The sulfate solubility was further improved by adding CaO tothe batch. The results, shown in table 11-3, indicate that an increase in CaO concentration decreases theformation of salt layer due to formation of CaSO4, which is a stable liquid. The sulfate solubilityinthe cruciblemelt is 0.35 wt% S03. When the same batch composition was fed to a melter, the sulfate solubilitywas 1.2 wt% S03 in the glass. The increase was attributed to the adjustments in the glass composition,presence of cold cap inthe melter, and alowermeltingtemperature inthe melter compared to crucible tests.Eventhoughthis studyisnotdirectlyapplicabletotheHanford glasses, itshowsthatlaboratoryand pilot-scalemeltertests could yield significantlydifferentresults. Conductinglaboratorytests followed bypilot-scaletestsis strongly recommended.
11.3 EFFECT OF MELTER CONDITIONS ON SULFUR SOLUBILITY
As discussed in section 11.2.1, redox conditions in the melt have significant effects on the sulfursolubility. Since sulfur segregation is a kinetic process, the melter conditions such as temperature and cold capconditions could have significant effects on the formation of gall. Secondary phases could form due toextremely reducing conditions inthe melt, which could precipitatehighlyconductive phases such as NiS, CuS,or FeS.
11.4 TANK WASTE REMEDIATION SYSTEM CONCERNS
At Hanford, LAW sulfate solubility is a major concern. In traditional HLW borosilicate glassesprocessed in the redox range of 0.01 to 0.5, the sulfur concentration should be < 0.25 wt% as S03 in glass.Laboratory studies have indicated that sulfur solubilityunder oxidizing conditions could be as high as 1.2 wt0/oS0 3 . Increasing the concentration of P205 and NBOs (alkali oxides) and decreasing the amount of A12 03
could improve the sulfur solubility in the melt. However, the kinetic behavior of sulfur segregation such as theformation of gall in the cold cap during the melting phase or the formation of secondary phases in the
11-7
Table 11-3. Formation of salt layer as a function of redox in glass containing 9 wt% CaO(Davis et al., 2000)*
Wt% Equivalent C I Salts Present After Melting? I Metal Present After Melting?
0.0 Yes-Minimum Amount None
0.1 None None
0.2 None None
0.3 None, but observed during melting Yes
0.4 None, but observed during melting Yes
*Davis, D.H., D.M. Bennert, and E. Nicaise. Control of batch redox to permit practical manufacturing of Femald POPTsurrogate-based glass. Ceramic Transactions of the Environmental Issues and Waste Management Technologies in the Ceramic
and Nuclear Industries VI, St. Louis, Missouri, April 30-May 3, 2000. Ceramic Transactions Volume 119. D. Spearing andV. Jain, eds. Westerville, OH: American Ceramic Society. In press.
presence of P205 needs further evaluation to ensure safe operations. Proprietary studies to improve sulfursolubility in melts without formation of gall or excessive volatilization of SO2 are being conducted at CatholicUniversity of America to support the Tank Waste Remediation System project. The results of these studiesare not published to date.
11.5 SUMMARY
Borosilicate glass-based waste form has a very limited sulfur solubility. Even though SO3 solubilityunder oxidized conditions is approximately 1 wt%, the S03 levels in the WVDP and DWPF glasses aremaintained below 0.25 wt% to avoid formation of an immiscible sodium sulfate phase in the melt becauseS03 solubilityin glass is astrong function of meltredox conditions. SO3 solubilitydecreases sharply as glassmelts become reducing. The S0 3 solubility limit is established based on an operating redox rangeof O.01 and 0.5. The sodium sulfate phase, usuallyreferred to as gall, is volatile, water soluble, and floats onthe surface of the melt. The formation of gall could result in undesirable consequences such as partition ofradionuclides into the gall phase, foam formation in the melter, excessive corrosion of refractories andAlloy 690 components in the melter, and disruption of heat balance in the melter (gall acts as an insulatinglayer on the surface of the melt). Studies have indicated that under oxidizing conditions, increasing theconcentration of P205 and NBOs (alkali oxides) and decreasing the amount of A1203 could improve sulfursolubilityin the melt. However, kinetic behavior of sulfur segregation such as formation of gall in the cold capduring melting phase or formation of secondary phases in the presence of P205 needs further evaluation toensure safe operations. Anyimprovementinincreasingthewasteloadingthroughbetterunderstanding ofthesulfur solubility could result in significant cost savings. Current plans at Hanford involve development of glasscompositions for LAW that can accommodate SO3 concentration in excess of 1 wt%.
11-8
11.6 REFERENCES
Bickford, D.F., and R.B. Diemer, Jr. Redox control of electric melters with complex feed compositions.Journal of Non-Crystalline Solids 84: 276-284. 1986.
Li, H., M.H. Langowski, and P.R. Hrma. Segregation of sulfate and phosphate inthevitrification of high-levelwastes. Proceedings of the International Symposium on Environmental Issues and WasteManagement Technologies in the Ceramic and Nuclear Industries. 97'h Annual Meeting of theAmerican Ceramic Society, Cincinnati, Ohio, May 1-5, 1995. Ceramic Transactions Volume 61.Westerville, OH: American Ceramic Society: 195-202. 1995.
Pabalan, R.T., M.S. Jarzemba, D.A. Pickett, N. Sridhar, J. Weldy, C.S. Brazel, J.T. Persyn, D.S. Moulton,J.P. Hsu, J. Erwin, T.A. Abrajano, Jr., and B. Li. Hanford Tank Waste Remediation SystemHigh-Level Waste Chemistry Manual. NUREG/CR-5751. CNWRA 97-008. Revision 2.San Antonio, TX: Center for Nuclear Waste Regulatory Analyses. 1999.
Schreiber, H.D., C.W. Schreiber, E.D. Sisk, and S. J. Kozak. Sulfursystematicsinmodel compositions fromWest Valley. Proceedings of the Environmental and Waste Management Issues in the CeramicIndustry II Symposium. 9 6th Annual Meeting of the American Ceramic Society, Indianapolis,Indiana, April 25-27, 1994. Ceramic Transactions Volume 45. Westerville, OH: AmericanCeramic Society: 349-358. 1994.
Stefanovsky, S.V., I.A. Ivanov, and A.N. Gulin. Aluminophosphate glasses with high sulfate content.Proceedings of the Materials Research Society Symposium. Symposium Proceedings 353.Pittsburgh, PA: Materials Research Society 101-106. 1995.
Sullivan, G. K. Sulfate segregation in vitrification of simulated Hanford nuclear waste. Proceedings of theInternational Symposium on Environmental Issues and Waste Management Technologies in theCeramic and Nuclear Industries. 9 7th Annual Meeting of the American Ceramic Society,Cincinnati, Ohio, May 1-5, 1995. Ceramic Transactions Volume 61. Westerville, OH: AmericanCeramic Society: 187-193. 1995.
11-9
12 CHEMICAL DURABILITY
When a glass comes in contact with an environment, such as flowing or stagnant groundwater, corrosivegases and vapors, or aqueous solutions, chemical reactions occur at the surface, and then spread to the wholeof the glass, depending on its composition, the pH of the solution, and the temperature of the environment.Newton (1985) provides an excellent historical review, dating back to 1660, of the chemical durability ofglasses. The modem understanding of chemical durability of glasses has developed in the last 30 yr. Thesestudies have focused onunderstandingtwo aspects of chemical durability ofnuclear waste glasses. First, theability to predict glass durability and produce glasses to meet specific leaching criteria based on short-termtests, and, secondly, the abilityto predict the long-term (onthe order of 10,000 yr or more) dissolution behaviorof glasses. In recent years, studies attempt to measure the durability of glasses developed for the disposalof high-level radioactive wastes. Various aspects of the chemical durability of glasses have been reviewedby Paul (1977), Jantzen (1992), Ellison et al. (1994), Bourcier (1994), and the U.S. Department of Energy(1994).
This chapter provides an overview of the current understanding of glass dissolution behavior. Since glasscorrosion is not an intrinsic property ofthe materials, the observed dissolution behavior depends on the testconditions and test methods. Several methods have been developed for accelerated testing to determine thecorrosion behavior of glasses. A few testmethods that formthe basis of current understanding are discussed,as well as relationships between glass composition and environment (pH, temperature, and nature of solution).A careful evaluation is required before drawing conclusions regarding the chemical durability of a glass basedsolely on data. In this report, the terms glass durability, corrosion of glass, or dissolution of glass areinterchangeably used to refer to glass corrosion.
12.1 CATION RELEASE SPECIFICATION
12.1.1 Production Specification
The WAPS require, at the time of shipment, the producer to demonstrate control of waste formproduction by comparing, either directly or indirectly, production samples to the EA benchmark glass. Theconsistency of the waste form must be demonstrated using PCT. For acceptance, the mean concentrationsof lithium, sodium, and boron in the leachate, after normalizing for the concentrations inthe glass, must eachbe less than those of the benchmark glass. One acceptable method of demonstrating that the acceptancecriteria are met would be to ensure that the mean PCT results for each waste type are at least two standarddeviations below the mean PCT of the EA glass.
12.1.2 Geologic Disposal Specification
The long-term performance of the HLW vitrified waste form and the overall performance of therepositoryare independent, thoughthere areno limits placed onthe radionuclide release from the glass wasteform. TheNRC, in Draft 10 CFR Part 63, has proposed a limit of 0.25 mSv (25 mrem) as the total effectivedose equivalent received in a single year, by the average member of the critical group, weighted by theprobability ofoccurrence as the overall system performance objective for the repository following permanentclosure. The review of repository performance is outside the scope of this report.
12-1
12.2 GLASS CORROSION MECHANISM
Glass reactions in aqueous environments are complex and depend on glass composition and contactsolution chemistry. Figure 12-1(a) shows a schematic of reactions occurring at various stages of the glassdissolution process. The processes involved in glass dissolution include ion-exchange, water diffusion,hydrolysis, and precipitation. Even though dissolution may start by simple ion-exchange or hydrolysis reaction,the simultaneous occurrence of the stated processes is not uncommon. The initial reaction on the glasssurface depends on the contact-solution chemistry. In contact with water at neutral pH, the dissolutionbehavior of a simple alkali silicate glass is controlled by the hydrolysis reaction expressed by Eq. (12-1).Alkaline pHinvolvesnetworkhydrolysis of Si-O-Si bonds, as shown byEqs. (12-2) and (12-3). InacidicpH,the reaction begins with an ion-exchange reaction between alkali ions (M+ ions) in the glass and the hydrogenions (hydronium ions, H30) in the solution, as shown by Eq. (12-4). In 1980, Enrsberger pointed outthat thefield intensity ofthe bare proton is so high that it cannot exist in a condensed phase, and the proton is probablyassociated with water molecules (H30). The interrelationship between reactions was described byPaul (1982). He stated that, in neutral solutions, water reacts with alkali ions at the NBO site to producehydroxyl bonds and release alkali ions into solution, which increases the pH of the solution, as shown byEq. (12-1). Next, the OH- ion disrupts the siloxane bonds, as shown by Eq. (12-2). The E Si-O-formed inreaction (12-2) further reacts with H20 producing an OH- ion [Eq. (12-3)], which is free to repeat thereaction in Eq. (12-2). The release of Si from the glass into the contact solution, as silicic acid, occurs whenBO bonds associated with Si are hydrolyzed, as shown by Eqs. (12-5) and (12-6).
3Si - 0 - M + H 2 0 --+Si - O - H + M+ + OH- (12-1)
Si - O - Si _ + OH- -_ Si -O - H + _ SiO- (12-2)
= Si - O + H 2 0 Si - - H + OH- (12-3)
Si-O- M+ H 30+-> Si-O- H+ M++H 2 0 (12-4)
= Si - OSi(OH)3 + OH- -= Si - 0- + Si(OH)4 (12-5)
= Si - 0 - Si(OH)3 + H2 0 - = Si - OH + Si(OH)4 (12-6)
As the corrosion front progresses inside the glass, more and more alkali ions are released from theglass into the solution and the solution, pH continues to rise. However, the increase in pH is counteracted bythe release of a weak acid (silicic acid). The release of silicic acid increases sharply at pH > 9, as shown infigure 12-2. The initial rate of glass corrosion depends on the mass transport rate ofH20 or H30+ to the alkalisite and removal of alkalis out of the glass into the contact solution. In acidic pH, the ion-exchange reactionprogresses at a much higher rate compared to the network hydrolysis reactions, as shown in figure 12-3.
As the reaction progresses, reaction layers are formed on the surface. The formation of reactionlayers on the surface of the glass in contact with leachate plays a significant role in determining the leachingbehavior of the glass components. Figure 12-1 (b) shows a schematic of the surface layers formed on thesurface of the glass. The outermost surface layer is called the precipitation layer, while the innermost layeris called the diffusion layer. The layer sandwiched between the diffusion and precipitation layers is called thegellayer. The formation of the diffusion layer occurs through ion-exchange reactions shown by Eqs. (12-1)
12-2
........................
............
............
............
........................
............
............
Glass begins toreact with water.
Gel layeri ,Difflusion layer
o Hydration and ion... . . . ... I... . exchange at the
surface.................................................
Secondary phases
:W * *, W, A. .... .. , :.. ............ ..................... .......*....
................................. .....
n ip
Diffusion layer thickensuntil rate of diffusion ofalkalis equals rate ofnetwork dissolution.
Diffusion layer maintainsconstant thickness asglass dissolves at steadystate. Secondary phasescontinue to grow.
(a)
Crystalline Precipitate Build-up.Y
Precpitated Layer,Amorphous and Crystalline
- Precipitates.a2a)
I-J8)
Gel Layer, Amorphousand Crystalline withSpecies Migration.
Diffusion Layer, ExhibitsExtensive Pitting Depletionof Soluble Elements (B, U, Na).Sometimes Water PenetrationWithout Depletion of Soluble Elements.
(b)
Figure 12-1. (a) Glass dissolution mechanism, (b) Schematic of surface layer on leached glass
12-3
40 -U)
'20
C3 6 9 12
pH
Figure 12-2. Effect of pH on the rate of extraction of silica from fused silica powder at 80 'C (Paul,1982)
2
0
Na2OCD)
Si2E
1 3 5 7 9 11 13pH
Figure 12-3. Effect of pH on the extraction of soda and silica from a Na2O 3SiO 2 glass at 35 ° C(Paul, 1982)
12-4
and (12-2), while the formation of the gel layer occurs through a hydrolysis reaction, shown by Eqs. (12-3)and (12-4). As the glass corrodes and its components are released into solution, the leachate may becomesupersaturated with some components, leading to precipitation ofthe secondary phases on the surface oftheglass or the walls of the test container.
The thickness of the gel layer remains constant if the pH of the contact solution does not change. Ifthe contact solution becomes more alkaline, the Si dissolution increases, which may reduce the thickness ofthe gel layer. In a more complex system containing more than one glass-forming oxide such as B203 or A1203
(depending on the glass composition), the corrosion mechanism is further complicated by the release rate ofvarious cations from the glass and their solubility in the contact solution.
The properties ofthe surface layer depend on the glass composition, contact solution chemistry, testparameters, and reaction time. The surface layers can provide a sink or reservoir for components in solutionand act as a physical barrier to the transport of reactants and corrosion products. In addition, the surfacelayers can influence the chemical affinity of the glass reactions.
Depending on the test conditions, surface layers could either increase or decrease the corrosion rate.Chick and Pederson (1984) showed that even though surface layers provide some transport barrier to glassreaction, the leaching behavior is controlled bythe contact solution. Conradt et al. (1985) showed that surfacelayers in glass corroded in brine solutions (pH=5.7) at 120 and 200 0C had no effect on the corrosion rate,but when the glass was tested in 0.1 M NaOH (pH=l 2), glass corrosion was completely controlled by surfacelayers. Grambow and Strachan (1984) tested simulated waste glass PNL 76-68 in deionized water and0.001 M MgCl2 solution and showed that the formation of surface layers in the MgCl2 solution dominatedthe corrosion behavior of the glass. The experimental evidence suggests that the surface layers provide abarrier to glass corrosion. However, the extent of this effect depends on the glass composition, characteristicsof reaction layers, contact solution, temperature, and test conditions.
Surface layers that act as sinks reduce the concentration of corrosion species in the solution andincrease the affinity for glass dissolution. The dissolution affinityis influenced by factors such as (Feng, 1994)
* Precipitated crystalline phases* Amorphous silica phase(s)* Gel layer* Bulk glass components
12.3 GLASS DURABILITY MEASUREMENTS
Glass durability is not an intrinsic material property. Glass durability is dictated by the parameterscontrolled during testing. Typically the test is either dominated by the solution chemistry or the glasschemistry. The tests conducted using a low surface area/volume (SANV) ratio are referred to asglass-dominated conditions. In this case, a monolithic sample is placed either in a large amount of water inclosed system or in a periodic or continuous flowing condition. The corrosion behavior is not influenced bythe glass components leached into the contact solution. The tests conducted using a high SAN ratio arereferred to as solution-dominated conditions. Inthis case, the contact solutions quickly attain saturation withthe glass components, and the test results are dominated by the glass components leached into the contactsolution. The glass components exceeding solubility limits in the solution often precipitate on the surface of
12-5
the glass and lead to the formation of secondary phases. Depending on their characteristics, these secondaryphases could either increase or decrease the dissolution rate. These tests are used to provide acceleratedcorrosion of glasses.
The corrosion of glass is usuallydetermined bymeasuringthe amounts of various glass componentsthat are released into the solution in contact with the glass. The components are released at different ratesdepending on the characteristics of the contact solution and the chemical composition of the glass. In alkaliborosilicate glasses, alkali and boron are released at the fastest rate and are often used to measure corrosion.Boron and alkali have high solubility in solution and are not incorporated into the secondary phases that areformed onthe surface ofthe glass. Boron is preferred over alkalis because, a boron release is more sensitiveto reaction kinetics such as temperature and pH. However, to study the reaction mechanisms, additionalcomponents such as Al, Si, and others involved in the reactions are monitored. Table 12-1 shows differenttechniques and protocols used in determining glass durability. The test methods are generally categorized asstatic tests and dynamic tests. In static tests, the leachate is not refreshed or replaced, while in dynamic tests,the leachate is continuously or periodically refreshed or replaced.
12.3.1 Static Tests
The most widely used static tests to compare durabilities of various glasses, and the accepted testmethod for determining corrosion behavior, are Materials Characterization Center (MCC)-1 and MCC-3(U.S. Department of Energy, 1982), and the PCT (ASTM C1285-97) durabilitytest. In MCC-1, amonolithicglass sample is used in test solutions such as deionized water, brine, and silicate solution. The referenceconditions include the SAN ratio of 10 m-l at 90 'C for 28 d. However, the conditions can be varied. Theglass corrosion is determined by analyzing the concentration of the glass components in the leachate. Thesamples for the test are prepared by fracturing, cutting, or grinding. The surface finish has a measurableeffect on the corrosion rate. The rougher sample has a larger surface area and, therefore, a higher corrosionrate. The test sample and the contact solution are placed in a closed vessel, usually perfluoroalkoxy (PFA)Teflon@, and placed in a oven equilibrated at the test temperature. The PCT method is an ASTM standard(Test Method C 1285-97), a modified version ofthe MCC-3 test, and is exclusively developed to monitor theperformance of the HLW glasses during production. This method uses a crushed sample, which providesa high SAN ratio (2,000 m' 1) instead of the monolithic sample in MCC- 1 (10 m '). The method details twoversions of the test. The PCT-A method uses a crushed glass specimen with a particle size distributionbetween -100 and +200 (0.149 and 0.074 mm) mesh, placed in an airtight 304L stainless-steel vesselcontaining deionized water, which is ten times the mass of the crushed sample. This provides a SAN of2,000 m-', and the test is conducted at90 'C for a fixed duration of 7 d. The normalized release concentrationfor element i, NC;, in the leachate can be calculated by Eq. (12-7).
NC, = Ci (12-7)Fj
where NCj is in g-glass/m3 , C, is the concentration of element i in solution in g/m3, and F; is the mass fractionof element i in glass. The PCT-B method allows variation in the SAN ratio, particle size, leachatecomposition, temperature and time. This method is used for investigating the effect of various parameters onglass corrosion. The test uses either stainless steel or PFA Teflone vessels. The high SAN ratio acceleratessaturation of glass components in the contact solution. A PCT can reach saturation in a few days while an
12-6
Table 12-1. Summary of leach test methods (U.S. Department of Energy, 1994)
Temperature 1at SampleName of Test (0 C) Leachant Flow Rate Description j Reference
Soxhlet 50-100 Distilled water 1.5 cm3 /min Plate, SA = 3 cm2 I
Modified soxhlet 35-100 Distilled water Variable Grains or plate 2SA = Variable
Hot-cell soxhiet 100 Distilled water 80 cm3/hr Beads/plates/chips 3SA = Variable
MCC-5 soxhlet 100 Distilled water -1.5 cm3 /min Plate, S - 4 cm2 4
Soxhlet (PNC) 70,100 Distilled water 60-225 cm3/hr Bar, SA 2 cm2 5
HIPSOL 100-300 Distilled water 100-900 cm 3/hr Powder or block 6(HT soxhlet) l
IAEA 25 Distilled water Periodic Cylinder with 7replacement exposed surfaces
ISO Buffer 23-100 Distilled water, Periodic Monoliths 7buffers, and replacementseawater
Powder (P1) 95-200 Distilled water Periodic Powder, 8replacement 100-200,um
Powder (P2) 40-100 Deionized water Daily Powder, 100-150 9replacement mesh
MCC-4 40, 70, 90 Distilled water 0.1-0.001 Plate, SA = 4 cm2 4and reference cm3/rinlgroundwater
Low-flow 25-90 Distilled water I cm3/wk Plate, SA = 3 cm2 10(radiotracers)
Dynamic 35-90 Distilled water 3-1,200 cm3/hr Grains or 11monoliths
Grain Titration 100 Distilled water Static Powder 12
Time-Dependent 20- -60 Buffered water Static Disc 10Method
MCC-1 40,70, 90 Distilled water Static Monolith 4and reference SAN = 10 m-'groundwater
MCC-2 110, 150, 190 Distilled water Static Monolith 4and reference SAN= 10 m 1
groundwater
12-7
Table 12-1. Summary of leach test methods (U.S. Department of Energy, 1994) (cont'd)
Temperature SampleName of Test (OC) Leachant Flow Rate Description Reference
MCC-3 40, 90, 110, 150, Distilled water Static (agitated) Crushed 4190 andreference (1) 149-175 m
groundwater (2) <45 ym
HTLT (CEC) 90, 110 150, 190 Distilled water Static SA = 4 cm2 13
Autoclave 150-200 Distilled water Static Beads, chips 14
(HM1) and brines
Autoclave (KfK) 100, 150,200, Distilled water Static Cylinders 15250 and brines SA = 20 or 5 cm2
Repository 25-90 Granite None (sampling Plate 10
Simulation equilibrated water equivalent to SA= 3 cm21cm3/no)
Waste/Water/ 98 Distilled water or Glass cube: 5Rock Leach granite water SA = 6 cm2
20 g granitepowder:250-710 ym, 60 cm2
water
MCC-14 25-250 Repository Static or Monoliths and 4groundwaters periodic powders
sampling
PCT 90 Deionized water Static Crushed 74-149 16yin
I Nakamura H., and S. Toshiro. Safety Research of High-Level Waste Management for the Period April 1982 toMarch 1983. Progress Report JAERI-M-83-076. Japan Atomic Energy Research Institute. 1983.
2 Hussain, M., and L. Kahl. Incorporation of precipitation from treatment of medium-level liquid radioactive waste
in glass matrix or ceramics together with high-level waste. Symposium Proceedings for the Ceramic in Nuclear
Waste Management. 1979.3 Commission of the European Communities. Testing and Evaluation of Solidified High-Level Waste Forms. Joint
Annual Progress Report EUR 10038. Commission of European Communities (CEC). 1983.4 U.S. Department of Energy. Nuclear Waste Material Handbook (Test Methods) Technical Information Center.
DOE/TIC-I 1400. Washington, DC: U.S. Department of Energy. 1981.5 Igarashi, H. et al. Leaching test of simulated HLW glass in the presence of rock. Atomic Energy Society of
Japan-1983 Annual Meeting. 1983.6 Senoo, M. et al. High-Pressure Soxhlet-Type Leachability Testing Device and Leaching Test of Simulated
High-Level Waste Glass at High Temperature. JAERI-M-8571. Japan Atomic Energy Research Institute. 1979.7 Draft International Standard Report. ISO/DIS-6961. 1979.8 Oversby, V.M., and A.E. Ringwood. Leaching studies on Synroc at 95 and 200 'C. Radioactive Waste
Management 23: 223. 1982.9 Reeve, K.D., et al. The development of testing of Synroc for high-level radioactive waste fixation. Symposium
Proceedingsfor the Waste Management 1: 249-266. 1981.
12-8
Table 12-1. Summary of leach test methods (U.S. Department of Energy, 1994) (cont'd)
Temperature |ll Sample lName of Test (OC) I Leachant Flow Rate Description Reference]
10 Van Iseghem, P., et al. Chemical Stability of Simulated HLW Forms in Contact with Clay Media. EUR-8424. CCC.1983.
11 Vaswani, G.A., et al. Development of Improved Leaching Techniques for Vitrified Radioactive Waste Products.BARC-1032. Bhabba Atomic Research Center. 1978.
12 Deutsches Institut for Normung. DIN Leach Test. Report 12111. Deutsches Institut for Normung. 1976.13 European Community Static High Temperature Test Summnary. CEC Report EUR 9772. 1985.14 Altenhein, F.K, et al. Scientific Basic International Symposium Proceedings for the Nuclear Waste Management:
363-370. 1981.15 Kahl, L., M.C. Ruiz-Lopez, J. Saidl, and T. Dippl. Preparation and Characterization of an Improved Borosilicate
Glass for Solidification of High-Level Radioactive Fission Product Solutions. Part 2: Characterization of theBorosilicate Glass Product. GP 98/12. Kernforschungszentrum Karlsruhe Report KFK-325 le. 1982.
16 American Society for Testing and Materials. Determining Chemical Durability of Nuclear, Hazardous, andMixed Waste Glasses: The Product Consistency Test. Annual Book of ASTM Standards. ASTM C 1285-97.Volume 12.01. 774-9 1. West Conshohocken, PA: American Society for Testing and Materials. 1997.
MCC-1 test, which is a solution-dominated system (low SA/V ratio), may take significant time to reachsaturation.
12.3.2 Dynamic Tests
The Soxhlet (Delage and Dussossoy, 1991), single-pass, flow-through (SPFT) test (McGrail et al.,1997) and periodic replacement tests are widely used as dynamic tests. The Soxhlet test can be conductedover atemperature range of 90 to 250 'C. The test apparatus consists of astainless steel reactor with aboilersupported by a condenser. The leaching vessel is located in the upper portion of the boiler. The waterevaporating fromthe boileris condensed in arefluxtube and is allowed to drip into the leaching vessel, whichcontains either a monolithic or crushed glass sample. The excess water from the leaching vessel overflowsinto the boiler. The corrosion rate is measured by periodically removing and analyzing the sample from theboiler. Since the leachate solution is recondensed, only deionized water can be used as a leachate solution.While the test allows studying awide temperature and pressure range, the testing apparatus is quite complex.The test is useful where high flow rates are required.
The SPFT test is used to measure forward reaction rate. In this test, leachate solution is passed ata constantrate, using aperistalticgpump, into acell containing amonolith or crushed glass. The solutionleavingthe cell is periodically collected and analyzed. The corrosion rate is determined by Eq. (12-8).
NRj = (C,,j- Cab) j (12-8)
where NR ii is the normalized release rate of i1 element at the h sampling, q3 is the flow rate at the timeperiodj in m3 /s, C His the concentration of component i at the time periodj, C,,b is the mean backgroundconcentration of component ifj is the mass fraction of component i in glass, and sj is the surface area overthe time period j- andj, m2. In glasses, the corrosion rate increases to a maximum value as the flow rate
12-9
increases. This maximum corrosion rate is called the forward reaction rate. In this test, conditions such asleachate composition, temperature, and pH can be varied to measure the reaction characteristics.
In periodic replenishment tests, leachate solutions are periodically removed from the ongoing statictests. The static tests, such as MCC-1, MCC-3, and PCT, can be used to conduct periodic replenishmenttests. The time of replenishment can be determined based on the aforementioned test conditions. Thecorrosion behavior of the glass depends on the amount of leachate replaced and the frequency ofreplacement.
The selection of atestmethod depends onthe characteristics ofthe systemunder study. For example,if the interactions between the glass and contact solution are important without the influence of solutionchemistry, tests such as MCC-1 or dynamic tests are recommended. If the effect of solution chemistry onthe glass corrosion behavior is important, then tests such as PCT are more useful. Most of the literature dataon glass durability are collected using either the MCC-1 or PCT method. Selecting the test method thatduplicates the performance requirements (or meets the anticipated environmental conditions during use) ofa glass is the key to conducting relevant tests.
12.4 PARAMETERS AFFECTING LEACHING BEHAVIOR
Test parameters such as SAN ratio, flow rate, and temperature for a given method could limit theusefulness of the data for the desired application.
12.4.1 Effect of Surface Area to Volume Ratio
Glass dissolution initiates on the surface. The greater the surface area, the greater the dissolution ofglass in the contact solution. However, the effect of surface area on glass dissolution is not linear. Ebert andBates (1993) studied the dissolution oftwo reference DWPF borosilicate glasses, SRL-13 1 A and SRL-202A,with SAN ratios of 10; 2,000; and 20,000 m' l. The results shown in figure 12-4 indicate that B release cannotbe linearly scaled as a function of SANVt. The higher, nonlinear release at the high SAN ratio has beenattributed either to the differences in the contact solution chemistry or mass transport effects. The changesin solution chemistry (i.e., the evolution of pH as a function of time) are shown in figure 12-5. Pedersonet al. (1983) hypothesized that thicker surface layers are formed in tests with low SAN ratio, which affectsthe mass transport to a greater degree compared to the thinner surface layers formed in high SAN tests,which is responsible for higher B release from high SAN samples. In addition, long-term tests can beaffected bythe changes in surface area. The release from the glass should be corrected to include the changein surface area of the glass with time due to pulverization from stresses, formation of secondary phases, ornormal dissolution of glass. If the effects of solution chemistry, surface area, and nature of surface layers isnot properly accounted for in the analysis, the extrapolation of short-term data to long-term behavior couldbe biased.
12.4.2 Effect of Flow Rates
In the dynamic tests, the leachate solution is continuouslyflushed at a given flow rate or periodicallyexchanged. Theincrease inions released fromrtheglass is counteracted bytheirremovalthrough continuousor periodic leachate exchanges. The maximum corrosion will occur if the flow rate is such that the
12-10
E
02
-11 2 3 4 5 6 7 8
log (SAIV-t) (d/m)
CL0
0
-11 2 3 4 5
log (SAN-t) (d/m)6 7 8
Figure 12-4. Concentration versus (SA/V)-time for release of B from (a) SRL-131 glass and(b) SRL-202 glass at 90 'C at (-) 10, (0) 2,000, and (*) 20,000 m-1 (Ebert and Bates, 1993)
12-11
13
11
l4p-4W'-*" I I I (a)0 200 400 600 800
Reaction Time (d)
13
12 -
11
10
9 _
(b)80 200 400 600 800
Reaction Time (d)
Figure 12-5. Leachate pH values (25 'C) versus reaction time for (a) SRL-131 glass and(b) SRL-202 glass reacted at 90 'C at (@) 10, (U) 2,000, and (*) 20,000 m-1 (Ebert and Bates,1993)
12-12
components released from the glass do not exceed the solubility limit. Figure 12-6 shows the normalized Naand Si release rates as a function of flow rate in SRL-131 glass. The curve shows two regions of leachingbehavior. In region 1 the leach rate is proportional to flow rate up to 1 ml/hr, while, in region 2 beyond 10mli/r, the leach rate is nearlyindependent of flow rate. Figure 12-7 shows the time dependence on corrosionat various flow rates. At high flow rates, the corrosion proceeds at a maximum rate that depends on theleachate composition, pH, and temperature but is independent of the flow rate. At a low flow rate, reactionrate decreases with time. The flow rate representing the maximum corrosion rate is referred to as a forwardreaction rate.
12.4.3 Effect of Temperature
In general, the temperature dependence of reaction rate is expressed through an Arrhenius equationof the type shown by Eq. (12-9).
k=Axe - (12-9)RT
where
k - reaction rate constant or normalized releaseA - preexponential constant,E - activation energy,T - temperatureR - gas constant.
Higher temperature is usually used as a method of accelerating the reactions, provided the reactionmechanism does not change over the range of temperature studied. Glass corrosion studies have beenconducted from ambienttemperatures to as highas 300 'C. However, most ofthe test methods are designedto study glass corrosion at 90 'C. Vernaz et al. (1988) studied the effect of temperature between 100 and300 'C onR7T7 French HLWborosilicate glass. The glass samples were corroded for various time intervals.The data shown in figure 12-8 indicate two distinct corrosion mechanisms between 100 and 250 'C. Theactivation energy for the corrosion process was about 30 kJ/mol, while for atemperature greater than 250 'C,the activation energy increased to 150 kJ/mol. The corroded samples showed an increase in the thickness ofthe secondary phase layer from 1.5 to 30 gim when the temperature increased from 100 to 250 'C. Thethickness of the secondary layer was almost 3 mm at 300 'C.
In addition, the leachate analysis showed that the effect of temperature on the release of variouselements from the glass was notthe same. Figure 12-9 shows the distribution of activation energies observedin various nuclear waste and natural glasses. The reported activation energy for glass corrosion ranges from22 to 150 kcallmol (U.S. Department of Energy, 1994). The wide range of activation energies is attributednot onlyto the complex nature ofthe glass corrosion mechanism but also to the effects of glass composition,contact solution chemistry, reaction time, and temperature. Therefore, temperature data should beextrapolated with care to assure the corrosion mechanisms do not change over the extrapolation range.
12-13
Linear Flow Rate, mLh-1
0.5 1.0 100.1 25 50I I I I I
Transition
IV
co.
a:
0)a,
a)T)a:
0z
3.0[ - Region 1 i
Solution Conc. Controlled
2.0 -
Region 2
Leach Rate Controlled
Na
Si
SRL 131 + 29.8% TDS 3A
I I I
1.0 _
I IOL
-1 .!5 -1.0 -0.5 0 0.5 1.0Log Flow Rate, mL-h 1
1.5 2.0
Figure 12-6. Normalized release rates(Hench et al., 1986)
6- SRLE SAN
,, 40 -D.I. %U,0
-j
a 30Ca
F 20~0a)N.FU10
0z
of sodium and silicon as a function of leachant flow rates
0 7 14 21Corrosion Time (d)
28
Figure 12-7. Normalized total mass loss versus leach time for various flow rates (Adiga et al.,1985)
12-14
----- -- - ------ -
Cl)
0 6 ------ 3 d------ E=150kJmor'Ca
E-=5OkJ moF 1
E E= 30 k J mol 1GI
'a E=24 k J moFcoE
0T
2 -
I I . _ I I I _
2 2.5 2 2.5 3
103/T -K 1
Figure 12-8. Normalized elemental mass loss versus temperature for different leaching durations(Vernaz et al., 1988)
12.4.4 Effect of Glass Composition
The chemical durability of a glass depends on its components. Because of the lack of a uniformtesting and measurement approach to study durability-composition relationships, the data from various studiesare difficult to compare quantitatively. Chemical durability-composition trends observed in various glasscompositions are summarized here.
The chemical durability behavior of fused silica glass powder at 80 'C as a function ofpH is shownin figure 12-2. Silica is amajor component of all glass-forming systems. The dataindicate that fused silica isfairly stable up to a pH of 9. Beyond pH 9, the network dissolution reactions, shown by Eqs. (12-5) and(12-6), are responsible for Si dissolution fromglass. Inglass systemsxNa 20 1OCaOo(90-x)Si0 2 containingalkali ions, Clark et al. (1976) showed that as Na concentration increases, the durability sharply decreases.In Na-Ca silicate glass, Das (1980) showed that the rates of Na and Si release were different as a functionof pH. The highest Na release rates were obtained in acidic conditions, while the highest Si release rates wereobtained in alkaline solutions, as shown in figure 12-3. The chemical durability also depends on the particularalkali ion present. Dilmore et al. (1 978a) showed that 15K20*1 OCaO.75 i0 2 glasses leach at a higher ratethan 15Na2 0* I OCaO*75Si0 2 glass and that mixed alkali glasses such as (15 -x)Na 2O-xK2 0 1 OCaO*75Si0 2are more durable than eitherNa or K end-members. The presence of more than one type of alkali in the glasssuppresses the alkali leaching from the glass. This behavior is called the mixed-alkali effect. Clark et al.(1976) showed thatthe addition ofa divalent cation, such as Ca, in glass significantly improves durability. Theconcentrations of Na and Si from 2ONa 2O 805i02 were 2.5 x 1O and 8.5 x IO' ppm after 12 hr. but in2ONa2O-l0CaO.70SiO 2 glass, the concentration was only 566 ppm after 9 d. Similarly, Smets and
12-15
Activation Energies (kJ/mol)
0 50 100 150
Etching/Surface Reaction
Precipitation Kinetics
Basalt Water Diffusion inGlasses |Obsidians/Teites
R7T7 Glass
IPNL76-68
|SRL Glasses
WasteGlass Japanese Glasses
Various Waste Glasses
CanadianGlasses
4 ) SM58, SAN60, UK209, PNL 76-68
0 50 100 150
Figure 12-9. Summary diagram indicating reported activation energies for individual reactionprocesses and overall activation energies for high-level waste glass studies (U.S. Department ofEnergy, 1994)
12-16
Tholen (1984) showed that in 20Na20 I OMO-70SiO 2 glass (where M = Ca, Mg, or Zn), glass durability asmeasured byNa remaining in the glass decreased with the presence of Ca, Mg, and Zn, in order. This studyalso indicated that the most drastic changes were observed at low concentrations of the divalent ions.
In alkali borosilicate glasses, depending on the alkali/boron ratio, boron is either incorporated in thestructure as tetrahedrally coordinated B04 species ortrigonallycoordinated B0 3 species. Adams and Evans(1978) studied durability/composition relationships by reacting the glasses in aNa2 O-B203-Si0 2 system at25 'C for 24 hr. Their study showed that the glass compositions that maximize the concentration of B0 4tetrahedra tend to have the highest durabilities. Bunker et al. (1986) reached a similar conclusion in theirdurability study on a Na2O-B203-Si0 2 glass system containing 60 mol% SiO2 as a function of NaO/B203ratio and as a function of SiO2 for Na2O/B203 = 1. They showed that in samples with Na2O/B203 ratios<<1, Na and B were preferentially leached from the glasses. However, in glasses with Na2O/B2 03 21, Naand B leached congruently in alkaline solutions. In addition, the glasses with Na2O/B203 = 1 were orders ofmagnitude more durable than glasses with a lower or higher Na2O/B 203 ratio.
In short-term durabilitytests (less than 1 hr) of an aLkali aluminosilicate [20Na 20xAl203 (1 -x)SiO2 ]system, Smets and Lommen (1982) showed the depth of the Na depletion decreased significantly with anincrease in the A1203 concentration. The greatest change in depth of Na depletion was obtained with smallamounts of added Al203, while for x = 5 or 10, the Na profile showed no measurable change. Similarly, forLi2O-A120 3-Si0 2 glasses, Dilmore et al. (1 978b) conducted durability tests at 100 °C for 41 d and showedthat as the A1203 concentration increases in the glass, Li ions released in the leachate solution decreased. Inalkali borosilicate glasses, depending on the alkali/alumina ratio, alumina is incorporated in the structure eitheras atetrahedrally coordinated A10 4 species (similarto B) oras a six-fold coordinated modifier inthe absenceof alkali ions in the glass structure. The improvement in durability is attributed to the formation of A10 4tetrahedra whose charge is balanced by alkali ions and forms an AlO4Na bond, which is much stronger thanNBO (-Si-O-Na) bond.
In alkali iron silicates, the fraction of Fe3+ ions that are stabilized by alkali ions in tetrahedracoordination depends on the redox [Fe2+/Fe3+ or Fe2+/Fe(1012')] ratio in the glass. As the redox ratio increasesor Fe3+ ions decrease, the number of FeO4 tetrahedra decreases. Feng et al. (1988) studied the effect of theredox ratio [Fe2+/Fe(IDal)] on durabilityusing the MCC-3 test onWV-205 simulated nuclear waste glass. Thedurability behavior is shown in figure 12-10. The normalized Na concentration increases sharply as thesamples are reduced. The decrease in durability is attributed to the fact that, as the redox ratio increases,more alkali ions attach as NBO (-Si-O-Na) to Si and fewer attach to the FeO4 (FeO4-Na) tetrahedra in theglass.
Feng et al. (1989) performed a studyonVWV-205 glass bysystematicallyvaryingmajor componentsbysmall concentrations and measuring durabilityusingthe MCC-3 test at 90 °C at various time intervals upto 180 din deionized water. Figures 12-11,12-12, and 12-13 showthe effect of small additions of SiO2,ZrO2and A1203, respectively, to WV-205 glass. The trends indicate that a significant increase in durability occurswith small additions of SiO2,ZrO2, and A120 3, followed by a plateau region with a constant high durability.Feng et al. (1989) explained the effect based on the deficiency of network formers. In the nondurable glassregion, the glasses can be considered deficient in network formers and, therefore, can be quickly attackedby water. Ellison et al. (1994) analyzed Feng et al. (1989) data and showed that WV-205 composition hada (Na+K+Li)/(B+Al) ratio of 1.8 and (Na+K+Li)/(B+Al+Fe) ratio of 1.3, which is significantly higher
12-17
18
E-a0
C/)
a)N
0z
,2150)EC 120
c.o
a)00 60)
310 20 40
Time (d)60
Figure 12-10. Leaching (MCC-3 test, 90 'C) of WV-205 glass as a function of glass redox state(Feng et al., 1988)
0
.4-C:a)
C0
000no
co
1000
500
Figure 12-11. Interpolated surface showing the dependence of MCC-3 boron concentration on testtime and amount of SiO2 added to WV-205 (Feng et al., 1989)
12-18
C0
.4-
a)C.)C0
C0oco
1000
-i- 500E
180
Figure 12-12. Interpolated surface showing the dependence of MCC-3 boron concentration on testtime and amount of ZrO2 added to WV-205 (Feng et al., 1989)
C0
Cca,WC0
000Mo
1000
500-
0)
80
Figure 12-13. Interpolated surface showing the dependence of MCC-3 boron concentration ontest time and amount of A1203 added to WV-205 (Feng et al., 1989)
12-19
than one. Thus, small additions of network formers were able to significantly improve the durability of theglasses.
Nuclear waste glasses contain alkalis (Li, Na, K, Cs), oxides, divalent (Ca, Mg, Sr, Ba) oxides,A1203, B203, Fe203, SiO2,ZrO2, and many minor components. The interactions between components makethe assessment of the composition/durability relationship challenging. Extensive work has been done toevaluate composition effects in nuclear waste glasses. The review of every waste glass system is beyond thescope ofthis report. Table 12-2 summarizes the effect of various components on glass durability from severalstudies. The composition/durability studies indicate that the greatest improvements in durabilityresult fromthe components that form the strongest bonds. For example, the durability of alkali silicate glass is improvedby adding A1203, B203, Fe203, ZrO2, or divalent cations (Ca, Mg, Zn, etc.) (shown in order of decreasingeffect). The mechanism in each case is a formation of bonds stronger than the alkali-NBO bond in simplealkali silicate glass. The addition ofAl20 3 , B203, or Fe2O3 removes one NBO bond and, to compensate, alkaliions attach to A104, B0 4, or FeO4 tetrahedras.
12.4.5 Effect of Contact Solution
The nature of the solution in contact with glass has a significant effect on chemical durability. Fengand Pegg (1994) leached simulated nuclear waste glass, WvV-205, in the presence of various alkali nitratesolutions at 90 'C. Figure 12-14 shows the effect of various alkali nitrates on the Si leach rate. The Si leachrates in various solutions were less than half those in deionized water. Reduced leaching in the presence ofa salt solution is attributed to the ion-exchange reaction shown in Eq. (12-10)
-Si - O-M2 + MI -> Si- OMI + M2(1-0
whereM, is the alkali cation in solution andM 2 is the alkali cation in glass. The ion-exchange reaction shownbyEq. (12-10) doesnotresultinnetworkhydrolysis as shownbyEqs. (12-3) and (12-4), orincreasethe pHof the solution. However, the reaction competes with and suppresses the network hydrolysis reaction, thuslowering the Si dissolution rate. Barkett et al. (1989) showed that simulated nuclear waste glasses leachedin Pacific Coast seawater have almost two orders of magnitude lower Si and Al release rates and one orderof magnitude lower B and alkali compared to the leach rates observed in deionized water. However, a recentstudy byWickert et al. (1999) showed a significant increase inthe dissolution rate of soda lime silicate glasswith increasing NaCl concentration in the leachate. Wickert et al. (1999) attributed the increase in theleaching rate to the replacement of H ions by alkali, ions on the surface sites, which facilitates the access ofaqueous medium in to Si-O-Si network.
Several studies have been conducted to determine the effect of container materials and corrosionproducts on glass dissolution behavior. McVay and Buckwalter (1983) and Bums et al. (1986) studied theeffect of metals on glass dissolution behavior. While the former showed higher glass dissolution in thepresence of ductile iron, the latter showed no significant effect on glass dissolution from 304L stainless steeland 409 and 430 ferrite steels. In addition, Bums et al. (1986) showed that A516 carbon steel had a significantdetrimental effect on glass dissolution. Inagaki et al. (1996), Bart et al. (1997), and Werme et al. (1983)studied the effect of magnetite on glass dissolution behavior. Magnetite is considered a primary corrosionproduct, and for many disposal systems these studies showed that glass dissolution is enhanced by thepresence of magnetite. Werme et al. (1983) also studied the effect of FeOOH and concluded that glass
12-20
Table 12-2. Leaching experimental results after varying composition (U.S. Department of Energy,1994)
Composition Change Durability Glass Composition Reference
(Ca, Mg, Zn) for Si Increases Simple glass 1
(Mg, Ca) for Si Increases Simple glass 2
Ca for Si Increases Simple glass 3
(Sr, Ba) for Si Decreases Simple glass 2
(Na, K, Li) for Si Decreases Simple glass 4
Na for Si Negligible Waste glass 5
Al for Si Increases Simple glass 6
Ca for (Na, K) Decreases Waste glass 5
Na for Al Negligible Waste glass 5
Na for K Variable Mixed alkali effect 7
Na for K Negligible No mixed alkali effect 5
Al for Fe Increases Waste glass 8
Fe3" for Zn Decreases Waste glass 9
Increase (Si, Al, Zr) Increases Waste glass 10, 11
Increase Al Increases Waste glass 12
Increase (Al, Cr, Si) Increases Waste glass, MCC-1 tests 13
Increase Al Decreases Waste glass, acid leachate 13
Increase B Decreases Waste glass 13
Increase B Increases Waste glass 10
Increase (Na, Li) Decreases Waste glass 14
Increase Alkali Decreases Waste glass 10, 11
Increase (Na, Ca) Decreases Waste glass, MCC-1 tests 13
Increase (Mg, Ca) Negligible Waste glass 10
Increase Ti Variable Waste glass 11
12-21
Table 12-2. Leaching experimental results after varying composition (U.S. Department of Energy,1994) (cont'd)
Composition Change Durability Glass Composition Reference
Increase Ti Increases Waste glass 13
Increase (Cu, Cr, Ni) Negligible Waste glass 11
Increase La Increases Waste glass 11
Increase Zn Decreases Waste glass, MCC-1 tests 13
Increase Zn Decreases Waste glass 11
Fe3 1 for Fe2" Increases Waste glass 10
Fe3+ for Fe2" Negligible Waste glass 8
I Smets, B.M., M.G.W. Tholen, and T.PA Lommen. The effect of divalent cations on the leaching kinetics of glass. J. Non-Crys. Sol. 65: 319-332. 1984.
2 Isard, J.O., and W. Muller. hifluence of alkaline earth ions on the corrosion of glasses. Phys. Chem. Glasses 27(2): 55-58.1986.
3 Rana, MA, and R.W. Douglas. The reaction between glass and water-Part I: Experimental methods and observations.Phys. Chem. Glasses 2(6): 179-195. 1961a.Rana, MA., and R.W. Douglas. The reaction between glass and water-Part 2: Discussion of the results. Phys. Chem.Glasses 2(6): 196-204. 1961b.
4 Douglas, R.W., and T.M.M. El-Shamy. Reactions of glasses with aqueous solutions. J. Am. Ceram. Soc. 50(1): 1-8. 1967.5 Tait, J., and D.L. Mandelosi The Chemical Durability ofAlkali Aluminosilicate Glasses. Atomic Energy of Canada Limited
Report AECL-7803. Pinawa Manitoba, Canada: Atomic Energy of Canada Limited. 1983.6 Smets, B.M.J., and T.PA. Lommen. The leaching of sodium aluminosilicate glasses studied by secondary ion mass
spectrometry. Phys. Chem. Glasses 23(3): 83-87.1982.7 Dilmore, M.F., D.E. Clark, and L.L. Hench. Chemical durability of Na2O-K20-CaO-SiO2 glasses. J. Am. Ceram. Soc. 61:
439-443. 1978.8 Van Iseghem, P., and R. de Batist. Corrosion mechanisms of simulated high level nuclear waste glasses in distilled water.
Riv.Della Staz. Sper. Vetro 5: 163-170. 1984.9 Nogues, J.L., and L.L. Hench. Effect of Fe2O3 /ZnO on two glass compositions for solidification of Swedish nuclear wastes.
Proceedings of the Materials Research Society Symposium. Symposium Proceedings 11. Pittsburgh, PA: Materials Research
Society: 273-278. 1982.10 Feng, X., I.L. Pegg, Y. Guo, Aa. Barkatt, P.B. Macedo, S.J. Cucinell, and S. Lai. Correlation between composition effects
on glass durability and the structural role of the constituent oxides. NucL. Technol. 85: 334-345. 1989.11 Macedo, P.B., S.M. Finger, AA Barkatt, I.L. Pegg, X. Feng, and W.P. Freeborn. Durability Testing with West Valley
Borosilicate Glass Composition-Phase II. West Valley Nuclear Services Topical Report DOE/NE/44139-48. West Valley,NY: West Valley Nuclear Services, Inc.. 1988.
12 Nogues, J.L., L.L. Hench, and J. Zarzycki. Comparative study of seven glasses for solidification of nuclear wastes.Proceedings of the Materials Research Society Symposium. Symposium Proceedings: 11. Pittsburgh, PA: Materials Research
Society: 211-218. 1982.13 Chick, LA, G.F. Piepel, G.B. Mellinger, R.P. May, W.J. Gray, and C.Q. Buckwalter. The Effects of Composition on
Properties in an I 1 -Component Nuclear Waste Glass System. Pacific Northwest Laboratory Report PNL-3188. Richland,
WA: Pacific Northwest National Laboratory. 1981.14 Diebold, F.E., and J.K. Bates. Glass-water vapor interaction. Adv. in Ceram. 20: 515-522. 1986.
12-22
2.5
2.0con | Ad- O-.1M KNOE 3E 1.5 0.1 M Ba(NO3)2
cc ------- 0. 1 M CsNO
CD)
00 50 100 150 200
Time (d)
Figure 12-14. The Si leach rates (mglm 2ld) of glass WV-205 in deionized water, 0.1 M KNO 3,0.1 M Ba(NO3)2 and 0.1 M CsNO3 solutions at neutral pH for up to 200 d of leaching
dissolution is higher in the presence of FeOOH than in the presence ofthe same amount ofmagnetite. Thesestudies clearly establish the effect ofmagnetite and FeOOH on enhancing glass dissolution. Pan et al. 1studiedthe behavior of simulated waste glass samples from WVDP and DWPF by subjecting them to long-termleaching tests in the presence of FeCl2 and FeCl 3 at 90 0C, to simulate a corroded waste package (WP)environment. Ferrous and ferric chlorides were selected because chloride is the primaryion responsible forWP corrosion. Results, as indicated in figure 12-15, showed substantiallyhigher initial normalized release forB and alkali, approximately a factor of 50 to 70 times greater than those in deionized water in 0.25-M FeCl3solutions. The initial leaching rate (figure 12-16) for B and alkali was found to be pH-dependent anddecreased as the leachate pH was increased. While the leach rate for Si did not show any significant changein the pH range studied, the leach rate for Al showed a minimum. The minimum for the leach rate of Aloccurred at different pH values. The study indicates that elements in the glass matrix are releasedincongruently.
As indicated by several studies, the contact solution has a significant effect on the glass dissolution.Studies, however, fail to indicate specific trends in the dissolution behavior in the presence of various saltsor components in the contact solution.
'Pan, Y.-M., V. Jain, M. Bogart, and P. Deshpande. Effect of iron chlorides on the dissolution behavior of simulatedhigh-level waste glasses. Ceramic Transactions of the Environmental Issues and Waste Management Technologies in the Ceramicand Nuclear Industries VI, St. Louis, Missouri, April 30-May 3, 2000. Ceramic Transactions Volume 119. D. Spearing andV. Jain, eds. Westerville, OH: American Ceramic Society. In press.
12-23
100.000
WVDP Reference 6 Glass
1 0.000 IMAI AB OK DLi ONa -Si]
13
I
1.000
0.1 0 0
0.010
0.00 1
.1W ~~ -- --------_ _ _ _ _ _ _ _ _ _ _
0 2 4 6
pH
8 10 12
Figure 12-15. Cumulative normalized leach concentration for boron versus time for the WestValley Demonstration Project (WVDP) glass in various solutions2
12.5 CHEMICAL DURABILITY MODELS
12.5.1 General Chemical Durability Models
Several attempts have been made to correlate chemical durability and glass composition. In thissection, proposed empirical and thermodynamic relationships for durability behavior are reviewed. Themodels discussed assume that the glass is homogeneous and crystal free. In addition, the models only applyto normalized release measured in static tests of fixed duration, and there is no correlation between datacollected using static tests and long-term durability. Static tests can only verify product consistency and cannot be regarded as an accurate measure of relative durability of a glass.
The free energy of hydration (FEH) model was initially developed by Paul (1 977) for simple alkalisilicate glasses and later extended by Jantzen (1992) to complex borosilicate-based waste glasses. FEHrelates the thermodynamic properties of glasses to their dissolution rates as measured in short-term,closed-system tests such as MCC-1 and PCT. FEH treats glasses as solids composed of mechanicalmixtures of silicate and oxide components as shown in table 12-3. The net free energy change (AGhyd) iscalculated as
AG hyd = EXx * (AG hyd)i (12-11)
2Jain, V., and Y.-M. Pan. High-Level waste glass dissolution in simulated internal waste package environments. Proceedingsof the Atalante 2000 Conference, October 24-26, 2000. Avignon, France. In press.
12-24
1000000
WVDP Reference 6 Glass
t 100000 A
10000 & -AA
32~~~~~~~~~~~ ~ 100 < -o-0.25M FeC12E00 a -^- 0.25M FeC13l
5 ~~~~~~~~~~~-- 0.002 5M FeC12|A- ---- 0.0025M FeC13
--- DIWater100 . -*---0.25 M HCI
0 1 0 20 30 40 50 60 70 80
Time, days
Figure 12-16. Normalized leach rate for various elements versus leachate pH after the firstsolution replacement for West Valley Demonstration Project (WVDP) glass3
where (AGhd)iis the free energychangeinkcal/mol ofthethermodynamicallymost stable hydration reactionof the structurally associated silicate and oxide of component i having mol fraction x,. The model assumesthat the choice of structural units is adequate and that free energy of mixing of the components in the glasscan be neglected. The original approach of Paul (1977) assumed that the silicate and borate components ofa glass hydrate to silicic and boric acid. However, for low-durability glasses where alkali released from theglass increases the pH of the solution >9.5, the solubilities of silica and borate rapidly increase. Therefore,additional contribution to the FEH should be included to account for silicic and boric acid dissolution atpH 9.5.The additional contribution is calculated by
A(AGhyd)=1.364 -log(1+ lo 10 + o-21994 f H2 SiO3 (12-12)~~h~dJ.[o~~~~+ 1 0 PH + 10 -2pH ) o
and
( 1 0 .+ 1 o2 L89 w o35.69A(AGhyd)=1.364 -log1 + +H +o-2H 1 -3 for H2 B0 3 (12-13)
'Jain, V., and Y.-M. Pan. High-Level waste glass dissolution in simulated internal waste package environments. Proceedingsof the Atalante 2000 Conference, October 24-26, 2000. Avignon, France. In press.
12-25
Table 12-3. Primary basis set of partial molar hydration free energies, AGi, for glass in oxidized pHregimes >7 (Jantzen et al., 1998)*
IAG 1 pHHydration Reactions kcalmole Range
A1203 + H120 -2AI0 2-(aq) + 2H (aq +37.68 7-14
AMO2 + 3H 20 -Am(OH)5-(aq) + H+(a4 +23.68 7-14
As2 03 + 3H2 0 + 02 -2HAs0 4-2(aq) + 4H+(aq) -33.65 7-11
B203 + 3H20 -3H 3BO3(aq) -10.43 7-10
{BaO + SiO2} + 2H2 0 -Ba(OH) 2 + H2SiO 3(a) -19.13 7-14
{CaO + SiO2} + 2H2 0 -Ca(OH) 2 + H2SiO3(aq) -9.74 7-14
{CdO + SiO2} + 2H20 -Cd(OH) 2 + H2SiO3(aq) +2.31 7-14
Ce 2O3 + 3H 20 2Ce(OH)3 -44.99 7-14
{CoO + SiO2} + 2 H20 + 1 02 Co(OH)3 + H2Si 3(aq) -2.33 7-142 4
Cr203 + 3H20- 2Cr(OH) 3 +11.95 7-14
(CS20 + SA) + 2H20 - 2Cs+(,q) + 20H-(aq) + H2SiO3(aq) -76.33 7-14
{Cu 20 + SiO2} + 102 + 3H20 -2Cu(OH) 2 + H2SiO3(aq) -18.85 7-142
{CUO + SiO2} + 2H20 -Cu(OH) 2 + H2SiO3(aq) +5.59 7-14
{FeO + SiO2} + 1 02 + 2 H20 -Fe(OH) 3 + H2Si0 3(aq) -17.28 7-144 2
Fe 20 3 + 3H20 2Fe(OH) 3 +14.56 7-14
{K20 + SiO2 } + 2H20 - 2K(aq) + 2OH-(aq) + H2SiO3(aq) -72.36 7-14
La203 + 3H20 -2La(OH) 3 -48.59 7-14
{Li20 + SiO2} + 2H20 - 2L+(aq) + 2 OH-(aq) + H2SiO3(aq) -19.99 7-14
{MgO + SiO2} + 2H20 Mg(OH) 2 + H 2Si0 3(aq) -2.52 7-14
{MnO + SiO2} + -02 + H20-MnO2 + H2SiO3(.q) -20.39 7-14
Mo 3 + H2 0D- 2H+(aq) + MoO 4 -2(aq) +16.46 7-14
12-26
Table 12-3. Primary basis set of partial molar hydration free energies, AG,, for glass in oxidized pHregimes >7 (Jantzen et al., 1998)* (cont'd)
1 AG, | pHHydraLon Reactions _kal/mole| Range
{Na2O + SiO 2} + 2H20 2Na(aq) + 20H-(4 + H2SiO3(.0 -49.04 7-14
Nd203 + 3H 20 -2Nd(OH) 3 -37.79 7-14
{NiO + SiO2} + 2H20 -Ni(OH) 2 + H2Si0 3(ao +4.42 7-14
P205 + 3H 20 - 2 HPO 4-2a( + 4H+ -26.55 7-12
{PbO + SiO 2 } + 2H20 - b Sf(a + H2 Si03 (. + +25.10 7-14
Pu0 2 + 3H 20 -Pu(OH) 5 -(aq) + H+(aq +30.75 7-14
{Rb2O + SiO2} + 2H2 0 - 2Rb-(aq) + 20H-(aq) + H2Si03(aq) -82.18 7-14
RU0 2 + 1 02 + H2 0 -Ru0 4-(.0 + 2H+(20 +57.30 7-10
Sb2O3 + 02 + H20 -2SbO3 (.0 + 2H+(.0 -37.56 7-14
SeO 2+ - 02 + H20 -SeO 42( ( + 2H+(2 -7.80 7-14
2
SiO2 H20 -H 2Si0 3(ao) +4.05 7-10
SnO2 + 2H20 -Sn(OH) 4 +10.16 7-12
{SrO + SiO2} + 2H20 -Sr(OH) 2 + H2SiO3(.0 -15.16 7-14
TcO2 + 102 + H 20 -TcO4-(.0 + 2H+(aq -2.04 7-14
TeO2 + 1 02 + H20 -TeO4 2(ZO + 2H+(ao +12.02 7-14
ThO2 + H20 Th(OH)4 +19.23 7-14
Jantzen, C.M., J.B. Pickett, KG. Brown, and T.B. Edwards. Method of Determining Glass Durability. Patent No. 5,846,278.Aiken, SC: Westinghouse Savannah River Company. 1988.
12-27
Linear relationships shown by Eqs. (12-12) and (12-13) were developed between the logarithmicnormalized release from the glass in g-glass/m2 and the calculated AGhyd including the effect of silicic andboric acid. Figure 12-17 shows normalized B and Si release formore than 300 datapoints collected using theMCC-1 test for 28 d at 90 0C as afunction corrected FEH. The relationship statistically showed that releaserate was better correlated with corrected FEH than uncorrected FEH. The more negative the FEH, the lessdurable the glass. Glasses, such as natural obsidians, tektites, basalts, silica, pyrex, window glass, and ancientRoman and Islamic glasses, were included in the study. The relationship between the Si and B normalizedrelease and the adjusted AGhyd is
log(NLsi) = -0.224 0AGhyd - 0.0448 r2 = 0.73 (12-14)
log(NLB) = -0. 2795AGhyd + 0.2147 r2 = 0.59 (12-15)
where Nu is the normalized release in g-glass/m2 for component i. Jantzen et al.4 refined the FEH model forpredicting durabilities ofthe HLW glasses produced at DWPF. The refined thermodynamic hydration energyreaction model (T]HERMO) calculates the FEH, AGf, as
AGf = AG + AGWA + AGSB>ll + AGSB>l2 (12-16)
where, AGp is the FEH, in kcal/I00 gm-glass, represented by Eq. (12-11), and the AGaiterms are thecontributions from the accelerated dissolution reactions. The AGaWA represents the weak acid (WA)disequilibrium effect, and AGaSB>1I and AGaSB>12 terms represent strong base effects at pH >11 and >12. TheAGaWA is sum of contributions from Eqs. (12-12) and (12-13). The relationship between the normalizedrelease (g/L) and the AGf was developed using the PCT-A (7 d at 90 0C) as
log[NLN3 (g / L)] = -0.0981AGf - 1.448 r2 = 0.88 (12-17)
log[NLB(g/L)]= -0.104AGf -1.535 r2 = 0.86 (12-18)
Note that the FEH and THERMO not only use different units for data analysis but also use differenttest methods. A step-by-step method for calculating FEH for the THERMO model is provided inJantzen et al.5
For vitrified HLW, Hrma et al (1994) conducted a multiyear statistically designed CVS tocharacterize the relationship between composition and properties. Table 5-1 shows target composition andupper and lowerlimits forvarious componentsinthe study. The target composition, HW-39-4, is based onthe proposed Hanford HLW glass composition for the NCAW waste (now known as envelope B/D waste).Glass durability was measured on 120 glasses using 28 d MCC-1 and 7 d PCT methods. Compositionaldependence on normalized release was determined by the first-order mixture models as
10
NR = bix, (12-19)
4Jantzen, C.M., J.B. Pickett, KG. Brown, and T.B. Edwards. Method of Determining Glass Durability. U.S. Patent
No. 5,846,27. Issued on December 8,1988.
5'bid.
12-28
10,000(a)
Log Si = -0.2240 Ghyd -0.0448
E
a)Co0-J
UDN
(E
-oE
0z03
1,0004-
1004-
Legendo ACTUAL
BASALT
o BOROSIL
* CHINESE
N EGYPTIAN
- FRENCH
* FRITS
* GLCERAM
o GRANITE
+ HIGHSI
° ISLAMIC
* LUNAR* MEDIEVAL
A MISC
O MIXALK
o OBSIDIAN
x ROMAN
o SRL
* TEKTITE
* VENETIAN
* WASTE
104-a
1 4-
0.111
I I
5 0 -5 -10 -15Free Energy of Hydration (kcafmol)
-20
10,000
(b)Log B =-0.2795 Ghyd -0.2147
EU)coU)0-Ja)co
-oCDN
E0zm
1,000-
100.-
a
E
10_-
Legend
EACTUAL
BASALT
*BOROSIL
I FRITS
:HIGHSI
OBSIDIAN
SRL
3TEKTITE
WASTE
1
0
0.110
I I
5 0 -5 -10 -15Free Energy of Hydration (kcaVmol)
-20
Figure 12-17. (a) Linear regression plots of the A Ghyd term versus silicon release to the leachantin a 28-d static leach test. The AGhyd term has been adjusted for pH changes, as discussed in thetext. (b) Linear regression plots of the AGhyd term versus boron release to the leachant in a 28-dstatic leach test. The J Ghd term has been adjusted for pH changes, as discussed in the text.
12-29
where
NR - the normalized release in g/m2
x; - mass fractionbi - regression coefficients for component i
The estimated regression coefficients for B, Si, Na, and Li are shown in tables 12-4 and 12-5 forMCC-1 and PCT data, respectively. For MCC-1 data, seven data points representing high B release wereremoved from the analysis. Their removal was attributed to nonlinear or interactive effects of components,which cannot be addressed byfirst-ordermixture models. Ther2 statistics indicate that-the first-order mixturemodel for MCC-1 data provides some predictive ability. Based on the regression coefficients provided intable 12-4, a component effects plot using normalized boron release was developed centered aroundcomposition HW-39-4 (figure 12-18). The analysis predicts thatnormalized B release is increased by Li20,
Na2O, and B203 (in the order written) and decreased by A1203 and SiO2. For the PCT data set, no data pointswere removed. The r2 values for the PCT data set were higher than the MCC-1 data set, indicating that thefirst-order mixture model for PCT data provides better predictive ability than MCC-1 data. Based on theregression coefficients provided in table 12-5, a component effects plot using normalized boron release wasdeveloped centered around composition HW-39-4, shown in figure 12-19. The analysis predicts thatnormalized B release is increased by Li20, B203, and Na2O and decreased by Al203 and SiO2. The MCC-1and PCT data were analyzed using first-order mixture model terms along with several second-order terms
Table 12-4. Regression coefficients for normalized release and pH using 28-d MCC-1 method(Hrma et al., 1994)
[ x. bb (SiB)b (Li) bH B) |
SiO2 0.9025 -0.0993 -0.2319 -0.0587 7.3144
B 2 03 6.8360 9.6800 9.4346 9.6671 9.1180
Na2 ° 7.1851 9.5454 9.1998 8.9623 17.8061
Li20 9.6783 11.8108 10.5425 12.2173 24.4490
CaO 2.0185 3.7572 4.7685 3.6144 13.3174
MgO 1.8500 5.1079 4.9844 5.0078 13.1459
Fe2O3 4.6516 5.7008 6.0434 5.7557 10.1817
A1203 -4.6186 -6.2911 -5.6794 -6.0434 4.0454
ZrO2 -2.1819 -0.6283 -0.2328 -0.7562 7.5917
Others 1.9548 3.6529 4.1125 3.4126 8.6750
#points 114 114 114 114 120
r2 -0.5996 0.6520 0.6525 0.6532 0.7487
12-30
Table 12-5. Regression coefficients for normalized release and pH using the 7-d productconsistency test method (Hrma et al., 1994)
|xi b, (Si)b, (B) b. (Li) i b (Na) b, (pH)
SiO2 -2.9671 -4.3173 -3.2278 -4.4118 8.1926
B2 0 3 -0.6148 11.9791 10.1496 9.4049 3.3332
Na 2 O 10.7384 17.6068 14.0017 19.4013 23.6171
Li 2 O 19.7370 22.5791 18.4159 19.0635 31.2448
CaO -6.0415 -8.7114 -5.3528 -1.9567 17.2056
MgO 2.9296 10.9210 7.1181 11.8228 15.3354
Fe 203 -4.2315 -3.2009 -4.5113 -4.0953 8.5939
A1203 -17.3377 -25.4071 -22.3095 -25.4294 5.3578
ZrO2 -10.8139 -10.5613 -10.0618 -11.4209 7.6111
Others -0.7297 0.1587 0.6181 0.6647 9.2689
# points 123 123 123 123 123
r _ 0.7377 0.8114 0.7905 0.8457 0.9092
selected using statistical variable selection techniques to improve the relationship between normalized releasefor B, Si, Na, and Li, and composition. The second-order mixture model provided a slightly better fit than thecorresponding first-order mixture models.
Tables 12-4 and 12-5 also show first-order mixture model regression coefficients for pH. The r2statistics for the MCC-1 and PCT pH models show thatthe first-order mixture model forPCT data providesbetter predictive ability than for MCC-1 data.
Feng and Barkalt (1988) developed astructural bond strength (SBS) that assumes the dissolution ofglass is controlled bythe breaking of network bonds such as Si-O-Si. The energy offormation that representsan average structural strength of the glass is calculated by Eq. (12-20).
AH = E XiAgl (12-20)
where AHis the total energy of formation, and AHi is the energy of formation for component i having molfraction xi in glass. The methodology of calculating the energy of formation of various glass components isdiscussed in chapter 6. The model was applied to various nuclear waste glasses from West Valley, SRL, andPNNL; commercial glasses, such as aluminosilicate and soda lime silica; and natural glasses such as basaltsand tektites. The tests were conducted using the modified MCC-3 test. The MCC-3 test is a closed-system
12-31
50
a,
ai)
coci)
ai)
N
coiE
z
00
C-)~00~
40 1
30 I
SiO 2
B2 03
.,;Na2O
s ~~~~~~~.7".. Li2 O .;'
A1203 /'Fe203
Zro 2 ~ Mg
Others . . g . .----
Fe2 033 CaOB2 03 L20\ ZrO2
Na20 2
- I I I I A20
20 1
101
-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16
Component wt% Change from HW-39-4 Value
Figure 12-18. Predicted component effects on MCC-1 release relative to the HW-39-4composition, based on the first-order mixture model using mass fractions fitted to the reduced dataset (Hrma et al., 1994)
12-32
101- Na 2O
IIE
ab)0a)
w(a(D
c)
.L(U0zI-
a-
(U0~
91-I
I8
/7 ISiO 2
/
6 _- If
5 _
4 _
3 -
2 1-Fe 2 O3 -- _
Others . . . . .
1 13�0� ......... �� ;Na 2 0 Li2 O 0 -iaO
2 I I I I I I I I I A1203l l l l l l l l l l l l l l l
0
-12 -10 -8 -6 -4 -2 0 2 4 6 8
Component wt% Change from HW-39-4 Value
10 12 14 16
Figure 12-19. Predicted component effects on PCT B release relative to the HW-39-4composition, based on the first-order mixture model using mass fractions (Hrma et al., 1996)
12-33
testusing a crushed sample. Linear relationship shown byEqs. (12-21) and (12-22) were developed betweenthe logarithmic normalized release from the glass and the calculated AH.
log(NLB 7-,Y) =-0.0193AH+5.2667 r2 = 0.705 (12-21)
log(NLB-28day) = -0.02865AH + 8.358 r2 = 0.723 (12-22)
Ellison et al. (1994) used an independent data set consisting of normalized release data using theMCC-1 test at 90 0C for 28 d for 300 glasses and evaluated prediction capabilities of the FEH, THERMO,CVS, and SBS models. The data set included waste glasses (alkali boroaluminosilicate, alkali borosilicate,alkali aluminosilicate, and alkali lime silicates), natural glasses, and historical glasses. Results showed that FEHprovided the best correlation between glass composition and durability among various models. FEH providedthe best results for alkali boroaluminosilicate glasses having high mass losses, but provided poor correlationfor highly durable glasses. For the SBS model, regression analysis showed the best correlation for alkaliboroaluminosilicate glasses (r2 = 0.80) and the worst for alkali borosilicate glasses (r2 = 0.05). Compared tothe FEH model, the SBS model provided a better correlation for predicting normalized release for highlydurable glasses. Neither the SBS nor the FEH models provided a good correlation for alkali borosilicateglasses.
Tovena et al. (1994) used normalized release data for 32 waste glass samples (6 R7T7 and 26 SON)to testthe FEH and SBS models. Both the FEH and SBS models provided satisfactory approximation ofinitialrelease rate. However, the models underpredicted the normalized release rate for glasses containing >15 wt%B203 with little A120 3 . This departure was attributed to the phase separation and complete dissolution oftheborate phase in the absence of A12 03-
The thermodynamic, structural, and empirical models reviewed inthis chapter were developed fora specific range of glass compositions and glass components. Even though the applicability ofthese modelscan be extended to other wastes, as shown by Ellison et al. (1994) and Tovena et al. (1994), the user iscautioned to perform a careful and rigorous evaluation prior to using any model for predicting normalizedrelease from the glass. In addition, the models predict glass performance under given test conditions at asingle time; therefore, the results cannot be used for predicting time-dependent durability behavior.
12.5.2 Long-Term Chemical Durability Models
The release of components from glass follows a complex nonlinear behavior as a function of time.The behavior is a combination of glass composition effects, changes in solution chemistrywith time, formationof secondary phases, and temperature. The long-term corrosion behavior can be divided into three distinctstages as reviewed by Ellison et al. (1994). In stage I, referred to as the short-term stage, the chemicalpotential gradient between the glass components and local environment is the steepest. The glass componentsare released into the local environment at a comparatively high rate. The soluble components, such as boronand alkalis, are released at a higher rate compared with components such as silica and aluminum oxide. Thishigher release rate results in the formation of a layer on the glass surface depleted of soluble components,compared with the bulk glass. This layer is often called the altered surface layer. In stage II, the intermediatestage, the corrosion rate decreases as the concentration of reaction products, particularly silica, increases inthe solutionthat is in close contact with the glass. In addition, the reaction products in the altered surface layer
12-34
reach saturation concentration oftheir crystalline phases and result in the formation ofsecondary phases, suchas zeolites and clays. In stage III, the long-term stage, glass corrosion rate is further affected because of thereprecipitation ofsecondaryphases that exceed solubilitylimits at the altered zone. Physical processes, suchas crystallization, cracking, or exfoliation ofthe altered surface layers, that occur in stage III, could influencethe glass corrosion rate, as well as the release and transport of colloids and radionuclides. The change indissolution rate also depends on the identity, distribution, and surface area ofthe secondary phases. In mostcases, the dissolution rate increases as a result of crystallization, exfoliation, and cracking of the alteredsurface layers. The transition from one stage to another is dependent on the glass composition and the localenvironment. A highly durable glass maytake months to years to reach stage II, whereas a nondurable glassmay reach stage II within hours or days.
The long-term dissolution models use a rate equation consistent with the transition state theory(Bourcier, 1994)
dn, = AnkrH aj-fI - e( aRT)J (12-23)
where
ni - is the number of mols of species i in solution released from the glassI -timeA - reactive surface area of glassvi - concentration of species i in the glasskr - rate coefficient for the rate-limiting reaction for glass dissolution
H a.n - the product of the activities (concentrations) of dissolved aqueous species, that
contributes to the activated complex of the rate-limiting microscopic dissolutionreaction
Af - reactionaffinity, defined asRTln(Q/K), where Q is the activity product andKis theequilibrium constant for the rate-determining glass dissolution reaction
a - a stoichiometric factor that relates the rate-controlling microscopic reaction to theoverall solid dissolution reaction (usually it is assumed a = 1)
R - gas constantT - temperature in kelvin.
The Eq. (12-23) implies that, at equilibrium, there is a reversible dissolution reaction. Since the glassis considered as thermodynamically metastable, the rate law cannot be applied to the overall glass dissolutionbutto some rate-limiting microscopic reversible reaction. Several parameters shown in Eq. (12-23) are notknown either from theory or experiments. Therefore, the rate equation is simplified to the form
dni = Antk(pH) l -(-) (12-24)dt K,
12-35
Equation (12-24) is commonly used for determining release rate. The parameters k, K, r, and a aredetermined by fitting experimental data. Grambow further simplified Eq. (12-24) to include only silica in theaffinity term and expressed glass dissolution rate as
RJ =k+I1-1 K(aq) )+Rfmal (12-25)
where
Rm - matrix dissolution rate
k, - rate coefficient
aSio2(aq - the activity of aqueous silica at the reacting surfaceK - glass saturation activityRf, 1 - residual rate after silica saturation is achieved
The parameters Kand k+ are regressed from experimental data. Geochemical modeling codes, suchas PHREEQE/GLASSOL, EQ3/6, DISSOL, REACT, and LIX1VER, use an expression shown byEqs. (12-24) and (12-25). Grambow'smodelhas been applied to several closed- and open-system dissolutiontests for a variety of glass compositions. In most cases, the model predicts observed trends and agrees withmeasured solution compositions to within a factor of two or three. The model, however, requires thatparameters such as K and k+ be determined for each glass composition. In addition, the release ofcomponents can only be modeled assuming congruent dissolution; the model does not provide amechanisticbasis for predicting the long-term dissolution rate of glasses. The long-term models have no capability topredict how the rate may change as the environmental parameters change during the lifetime.
12.6 GLASS BEHAVIOR IN THE REPOSITORY ENVIRONMENT
Several long-term HLW glass corrosion studies have been conducted in the previous 20 yr onsimulated glasses doped with plausible radionuclides, and fully radioactive glasses. Drip tests, designed tosimulate slow flow throughthe breached canisters, have been used by Fortner and Bates (1996) and Fortneret al. (1997) to study the long-term performance of actinide-doped WVDP and DWPF HLW glasses. Thelong-term PCT-B, designed to simulate fully immersed conditions, has been used by Ebert and Tam (1997)to studythe long-term performance of DWPF glasses. In addition, vapor hydrationtests (VHT), designed toreplicate a natural alteration process, are used by Luo et al. (1997) to compare the dissolution behavior ofDWPF glasses with that of naturally occurring basalt glasses.
The dissolution rate of the HLW glass decreases as the groundwater environment in close contactwith the WPs becomes saturated with glass matrix components, such as silica. Even though the glasscorrosion studies discussed previously confirmed that net dissolution rate decreases as the surroundingenvironment becomes rich in HLW glass matrix components, the drip test studies show a steep increase inradionuclide release rate for Pu and Am after 400 wk. The steep increase in radionuclide release rate wasattributed to the spalling of radionuclide-containing colloids from the exposed HLW glass surface. The HLWglass corrosion models currently proposed do not account for such excursions in corrosion behavior, althoughthey can have a significant effect on radionuclide release.
12-36
The concentration of silica in the near-field environment may affect the degradation of SNF WPs.If there is abundant silica during SNF dissolution, uranosilicates may eventually form. If silicais depleted andgroundwater contact with WPs is limited, schoepite may fonn. Because the retention factors ofradionuclidesin uranosilicates and schoepite may be different, the silica release from HLW glass corrosion may affect therelease behavior ofradionuclides from SNF dissolution. The effects of HLW glass corrosion on radionuclideretention in secondary minerals during SNF dissolution should be considered.
Long-term corrosion studies ofHLW glasses indicate formation of secondary phases on the exposedsurface ofthe HLW glasses. This process is dependent onthe external environment. Long-term PCTs in J- 13water show formation of clay, Ca-phosphate, and (Th, U, and Ca) titanate as secondary phases (Bates, 1998),whereas, the VHTs show accumulation of clay, zeolites, Ca-silicates, weeksite, and K-feldspar as secondaryphases (Bates, 1998). The formation of different phases under diverse test conditions is attributed to varyingsolution chemistries. The formation of secondary phases may also be influenced bythe corroding containermaterials. Secondary minerals play an important role in radionuclide release because they can incorporatelow-solubilityradionuclides, such as Pu and Am, and control their solubilitylimits. Theymay also actto blockthe reactive surface area of the primary phase.
It is important that the long-term radionuclide release rate in the corrosion models includes theinfluence of stage III behavior. Experimental data on the formation of secondary phases under anticipatedrepository conditions are necessary if the contribution of the HLW glass to the estimated receptor dose issignificant. If the models are simply based on experimental dissolution data for stages I or II that exhibitsignificant retention of radionuclides inthe secondaryphases, evaluation ofthe long-term radionuclide releaserates could be erroneous.
Natural analog studies, coupled with experimental data and geochemical modeling, provide othermethods of gaining confidence in predicting long-term corrosion behavior of glasses. Natural analog studiesare useful in evaluating the merits of extrapolating short-term experiments to longer time frames. Severalnatural glasses, especiallybasalt, have compositions comparable to the HLW glasses and have been subjectedto conditions similarto those expected inthe proposedYuccaMountainrepository(Ewing et al., 1998). Thecharacterization of secondary phases formed on these natural glasses can provide insights into the long-termdissolution behavior of HLW glasses.
A recent study by Luo et al. (1997) compared the formation of secondary phases in the naturallyoccurring Hawaiian basaltic glasses with the results of VHTs conducted for 7 yr on simulated basaltic andHLW borosilicate glasses. Luo et al. (1997) concluded that secondary phases formed on both simulatednatural glasses and HLW borosilicate glasses were similar to secondary phases observed in naturallyoccurring basaltic glasses, and VHTs could be used to simulate naturally occurring conditions.
Field data onnaturallyoccurring glasses, combined with experimental data and models on dissolutionof HLW glasses, could be useful to demonstrate that long-term dissolution behavior under repositoryconditions can be represented by extrapolation of results from short-term laboratorytests. Such data can beimportant to supplement and support the validity ofthe existing glass-dissolution data generally obtained byshort-term experiments.
12-37
12.7 TANK WASTE REMEDIATION SYSTEM CONCERNS
Chemical durability is an important specification for the vitrified HLW. The target glass compositionshould be designed to meet the specification and to accommodate variabilityin leachate concentrations causedby composition variability. A robust glass composition enveloping composition variability should be able tomeet the WAPS.
12.8 SUMMARY
When a glass comes in contact with an environment, such as flowing or stagnant groundwater,corrosive gases and vapors, or aqueous solutions, chemical reactions occur at the surface, whichthen spreadto the whole of the glass, depending on its composition, the pH of the solution, and the temperature of theenvironment. Several studies have focused on understanding two aspects of nuclear waste glasses. First, theabilityto produce glasses to meet production specifications based on short-termtests and second, predictingthe long-term dissolution behavior of glasses on the of 10,000 yr or more. Several methods have beendeveloped to perform accelerated testing to determine corrosion behavior of glasses. Since glass corrosionis not an intrinsic property ofthe materials, the observed dissolution behavior depends on the test conditionsand test methods. A careful evaluation is required in drawing conclusions regarding the chemical durabilityof a glass based solely on data. A few test methods that form the basis of current understanding are
discussed, including relationships between glass composition and environment (pH, temperature, and natureof solution) as a function of time. The processes involved in glass corrosion include ion-exchange, waterdiffusion, hydrolysis, and precipitation. Even though dissolution may start by simple ion-exchange or thehydrolysis reaction, the simultaneous occurrence of stated processes is not uncommon.
The initial reaction on the glass surface is dominated by the initial pH ofthe contact solution. As thereaction progresses, reaction layers are formed on the surface. The formation of reaction layers on thesurface of the glass in contact with leachate plays a significant role in determining the leaching behavior ofthe glass components. As the glass corrodes and its components are released in the solution, the leachate maybecome supersaturated with respect to some components, leading to precipitation of the secondary phaseson the surface of the glass or the walls of the test container. The thickness of the gel layer remains constantprovided the pH ofthe contact solution does not change. If the contact solution becomes more alkaline, the
Si dissolution increases, which may reduce the thickness of the gel layer. In a complex system containingmore than one glass forming oxides, such as B203 and A1203 (depending on the glass composition), thecorrosion mechanism is further complicated by the release rate of various cations from the glass and theirsolubility in the contact solution. The properties of the surface layer depend on glass composition, contact
solution chemistry, test parameters, and reaction time. The surface layers can provide a sink or reservoir forcomponents in solution and act as a physical barrier to the transport of reactants and corrosion products. Inaddition, surface layers can influence the chemical affinity of the glass reactions.
The selection of a test method depends on the characteristics of the system to be studied. Forexample, if the interactions between the glass and contact solution are important without the influence of
solution chemistry, tests such as MCC-1 or dynamic tests are recommended. If the effect of solutionchemistry on the glass corrosion behavior is important, tests such as PCT are more useful. Most of theliterature data on glass durability was collected using either the MCC-1 or the PCT method. Test parameters
12-38
such as SAN ratio, flow rate, and temperature for a given method could limit the usefulness of the data forthe desired application.
The chemical durability of a glass depends on its components. Because of the lack of a uniformtesting and measurement approach to studydurability-compositionrelationships, data from various studies aredifficult to compare quantitatively. The composition/durability studies indicate that the greatest improvementsin durabilityresult from adding components that form the strongest bonds. For example, the durability ofalkalisilicate glass is improved by adding A1203, B203, Fe203, ZrO2, or divalent cations (Ca, Mg, Zn). Themechanism in each case is formation of bonds stronger than the alkali-NBO bond in simple alkali silicateglass. In addition, the nature ofthe solution in contact with glass has a significant effect on chemical durability.
Several attempts have been made to correlate chemical durability and glass composition. The modelsassume that the glass is homogeneous and crystal free. The thermodynamic, structural, and empirical modelsdiscussed in this report were developed for a specific range of glass compositions and glass components.Even though the applicability of these models can be extended to other wastes, the user is cautioned toperform a careful and rigorous evaluation prior to using anymodel for predicting normalized release from theglass. In addition, the models predict glass performance under given test conditions at a single time; therefore,the results cannot be used for predicting time-dependent durabilitybehavior. The long-term dissolution modelsuse rate equations consistent with the transition state theory. Grambow's model has been applied to severalclosed- and open-system dissolution tests for a variety of glass compositions. In most cases, the modelpredicts observed trends and agrees with the measured solution compositions to within a factor of two orthree. However, the model does not provide a mechanistic basis for predicting the long-term dissolution rateof glasses. The long-term models have no capabilityto predict how the rate may change as the environmentalparameters change during the performance period for the glass.
12.9 REFERENCES
Adams, P.B., and D.L. Evans. Chemical durability of borate glasses. Borate Glasses: Structure,Properties, Applications. New York: Plenum Press. 1978.
Adiga, RB., E.P. Akomer, and D.E. Clark. Effects of flow parameters on the leaching of nuclear wasteglass. Ceramic Transactions of the Materials Processing and Design: Grain-Boundary-Controlled Properties of Fine Ceramics II, Aichi, Japan. Ceramic Transactions Volume 44. K.Niihara, K Ishizaki, and M. Isotani, eds. Pittsburgh, PA: Materials Research Society: 45-54. 1985.
American Society for Testing and Materials. Determining Chemical Durability of Nuclear, Hazardous,and Mixed Waste Glasses: The Product Consistency Test. Annual Book of ASTM Standards.ASTM C 1,285-1,297. Ceramic Transactions Volume 12.01. 774-91. West Conshohocken, PA:American Society for Testing and Materials. 1999.
Barkatt, A.A., E.E. Saad, R. Adiga, W. Sousanpour, A.L. Barkett, M.A. Adel-Hadadi, J.A. O'Keefe, andS. Alterescu. Leaching of natural and nuclear waste glasses in sea water. AppI Geochem4: 593-603. 1989.
12-39
Bart, G., H. U. Zwicky, E.T. Aemne, T.H. Graber, D. Z'Berg, and M. Tokiwai. Borosilicate glass corrosionin the presence of steel corrosion products. Proceedings of the 99g Annual Meeting of theAmerican Ceramic Society, Cincinnati, Ohio. Ceramic Transactions Volume 84. W. Wong-Nu,U. Balachandran, and A. Bhalla, eds. Westerville, OH: Materials Research Society: 459-470.1997.
Bates, J.K. Distributionand stabilityofsecondaryphases. Workshop onPreliminaryInterpretations WasteForm Degradation and Radionuclide Mobilization Expert Elicitation Project. San Francisco,CA. 1998.
Bourcier, W.L. Critical Review of Glass Performance Modeling. ANL-94/17. Argonne, IL: ArgonneNational Laboratory. 1994.
Bunker, B.C., G.W. Arnold, D.E. Day, and P.J. Bray. The effect of molecular structure onborosilicate glassleaching. Journal of Non-Crystalline Solids 87: 226-253. 1986.
Bums, D.B., B.H. Upton, and G.G. Wicks. Interactions of SRP waste glass with potential canister andoverpack metals. Journal of Non-Crystalline Solids 84: 25 8-267. 1986.
Chick, L.A., and L.R. Pederson. The relationship between reaction layerthickness and leach rate fornuclearwaste glasses. Proceedings of the Materials Research Society Conference. SymposiumProceedings 26. Pittsburgh, PA: Materials Research Society: 635-642. 1984.
Clark D.E., M.F. Dilmore, E.C. Ethridge, and L.L. Hench. Aqueous corrosion of soda-lime-silica glass.Journal of American Ceramic Society 59: 62-65. 1976.
Conradt, R, H. Roggendorf, and H. Scholze. Investigations of the role of surface layers in HLW glassleaching. Proceedings of the Materials Research Society Conference. SymposiumProceedings 50. Pittsburgh, PA: Materials Research Society: 203-210. 1985.
Das, C.R. Diffusion controlled attack of glass surfaces byaqueous solutions. JournalofAmerican CeramicSociety 63: 160-165. 1980.
Delage, F., and J.L. Dussossoy. R7T7 Glass dissolution measurements using high-temperature soxhlet device.Proceedings of the Materials Research Society Conference. Symposium Proceedings 212.Pittsburgh, PA: Materials Research Society: 41-47. 1991.
Dilmore, M.F., D.E. Clark, and L.L. Hench. Chemical durability of Na2O-K 2O-CaO-SiO2 glasses.Journal of the American Ceramic Society 61: 439-443. 1 978a.
Dilmore, M.F., D. E. Clark, and L.L. Hench. Aqueous Corrosion of Lithia-Alumina-Silicate Glasses.Journal of the American Ceramic Society 57: 339-353. 1978b.
Ebert, W.L. and J.K. Bates. A comparison of glass reaction at high and low SANV. NuclearTechnology 104(3): 372-384. 1993.
12-40
Ebert, W.L., and S.W. Tam. Dissolution rates of DWPF glasses from long-term PCT. Proceedings of theMaterials Research Society Conference. Symposium Proceedings 465. Pittsburgh, PA: MaterialsResearch Society: 149-156. 1997.
Ellison, A.J.G., J.J. Mazar, and W.L. Ebert. Effect of Glass Composition on Waste Form Durability: ACritical Review. ANL-94/28. Argonne, IL: Argonne National Laboratory. 1994.
Ewing, R.C., W. Lutze, and A. Abdelouas. Natural glasses and the 'verification' ofthe long-term durabilityof nuclear waste glasses. Proceedings of the XVIII International Congress on Glass (CD RomEdition). Westerville, OH: American Ceramic Society. 1998.
Feng, X., and A. A. Barkatt. Structural thermodynamic model for durability and viscosity of nuclear wasteglasses. Proceedings of the Materials Research Society Conference. Symposium Proceedings112. Pittsburgh, PA: Materials Research Society 543-554. 1988.
Feng, X, A.A. Barkett, and T. Jiang. Systematic composition studies on the durability of waste glass WV205.Proceedings of the Materials Research Society Conference. Symposium Proceedings 112.Pittsburgh, PA: Materials Research Society: 673-683. 1988.
Feng, X., I.L. Pegg, A.A. Barkatt, P.B. Macedo, S.J. Cucinel, and S.Lai. Correlation between compositioneffects on glass durability and structural role of the constituent oxides. Nuclear Technology85: 334-345. 1989.
Feng, X. Surface layer effects on waste glass corrosion. Proceedings of the Materials Research SocietyConference. Proceedings 333. Pittsburgh, PA: Materials Research Society: Symposium: 55-67. 1994.
Feng, X., and I.L. Pegg. Effects of salt solutions on glass dissolution. Physics and Chemistry Glasses35(2). England. 1994.
Fortner, J.A., and J.K. Bates. Long-termresults fromunsaturated durabilitytesting of actinide-doped DWPFand WVDP waste glasses. Proceedings for the Materials Research Society Conference.Symposium Proceedings 412. Pittsburgh, PA: Materials Research Society: 205-211. 1996.
Fortner, J.A., S.F. Wolf, E.C. Buck, C.J. Mertz, and J.K. Bates. Solution-borne colloids from drip tests usingactinide-doped and fully-radioactive waste. Proceedings for the Materials Research SocietyConference. Symposium Proceedings 465. Pittsburgh, PA: Materials Research Society: 165-172.1997.
Grambow, B., and D.M. Strachan. Leach testing of waste glasses under near-saturation conditions.Proceedings for the Materials Research Society Conference. Symposium Proceedings 26.Pittsburgh, PA: Materials Research Society: 623-634. 1984.
Hench, L.L., D.E. Clark, and A.B. Harker. Review: Nuclear waste solids. JournalofMaterialScience 21:1,457-1,478. 1986.
12-41
Hrma, P.R., G.F. Piepel, M.J. Schweiger, D.E. Smith, D.-S. Kim, P.E. Redgate, J.D. Vienna, C.A. LoPresti,D.B. Simpson, D.K. Peeler, and M.H. Langowski. Property/Composition Relationships forHanford High-Level Waste Glasses Melting at 1,150 rC. Volume 1 (Chapters 1-11).PNL-1 0359. Richland, WA: Pacific Northwest National Laboratory. 1994.
Inagaki, Y., A. Ogata, H. Furuya, K. Idemistu, T. Banba, and T. Maeda. Effects ofredox condition on wasteglass corrosion in the presence of magnetite. Proceedings for the Materials Research SocietyConference. Symposium Proceedings 412. Pittsburgh, PA: Materials Research Society: 257-264.1996.
Jantzen, C.M. Nuclear waste glass durability: I, predicting environmental response from thermodynamic(Pourbaix) diagrams. Journal of the American Ceramic Society 75(9): 2,433-2,448. 1992.
Luo, J.S.,W.L. Ebert, J.J. Mazer, and J.K. Bates. Dissolution rates of DWPF glasses from long-term PCT.Proceedings of the Materials Research Society Conference. Symposium Proceedings 465.Pittsburgh, PA: Materials Research Society: 157-163. 1997.
McGrail, B.P., W.L. Ebert, A.J. Bakel, and D.K. Peeler. Measurement of kinetic rate law parameters on aNa-Ca-Al borosilicate glass for low-activity waste. Journal of NuclearMaterials 249: 175-189.1997.
McVay G.L., and C.Q. Buckwalter. Effect of iron on waste-glass leaching. Journal of the AmericanCeramic Society 66(3): 170-174. 1983.
Newton, R.G. The durability of glass-A review. Glass Technology 26(1): 21-28. 1985.
Paul, A. Chemistry of glass-A thermodynamic approach. Journal of Material Science 12: 2,246-2,268.1977.
Paul, A. Chemistry of Glass. New York: Chapman and Hall, Ltd. 1982.
Pederson, L.R., C.Q. Buckwalter, G.L. McVay, and B.L. Riddle. Glass surface areato solutionvolume ratioand its implication to accelerated leach testing. Proceedings of the Materials Research SocietyConference. SymposiumProceedings 15. Pittsburgh, PA: Materials Research Society: 47-54.1983.
Smets, B.M., and T.P.A. Lommen. The leaching of sodium aluminosilicate glasses studied bysecondaryionmass spectroscopy. Physics and Chemistry, Glasses 23: 83-87. 1982.
Smets, B.M., and M.G.W. Tholen. Leaching of glasses with molar compositions20Na2O. OCaO-xAl20 3.(1 -x)SiO2 . Journal of American Ceramic Society 67: 281-284. 1984.
Smets, B.M.J., and M.G.W. Tholen. The pH dependence of the aqueous corrosion of glass. Physicsand Chemistry, Glasses 26: 60-63. 1985.
12-42
Tovena, I., T. Advocat, D. Ghaleb, E. Vemaz, and F. Larche. Thermodynamic and structural modelscompared with the initial dissolution rates of "SON" glass samples. Proceedingsfor the MaterialsResearch Society Conference. Symposium Proceedings 333. Pittsburgh, PA: Materials ResearchSociety: 595-602. 1994.
U.S. Department of Energy. Nuclear Waste Materials Handbook-Test Methods. DOE/TIC-1 1400.Richland, WA: Pacific Northwest National Laboratory. 1982.
U.S. Department of Energy. High-Level Waste Borosilicate Glass: A Compendium of CorrosionCharacteristics. (Volumes 1, 2, and 3). DOE-EM-0177. Washington, DC: Office of WasteManagement, U.S. Department of Energy. 1994.
Vemaz, E.Y., J.L. Dussossoy, and S. Fillet. Temperature dependence of R7T7 nuclear waste glass alterationmechanisms. Proceedings of the Materials Research Society Symposium. SymposiumProceedings 112. Pittsburgh, PA: Materials Research Society: 555-563. 1988.
Werme, L., I.K. Bjomer, G. Bart, H.U. Zwicky, B. Grambow, W. Lutze, R.C. Ewing, and C. Magrabi.Chemical corrosion of highly radioactive borosilicate nuclear waste glass under simulatedrepository conditions. Journal of Materials Research 5(5): 1,130-1,146. 1983.
Wickert, C.L., A.E. Vieira, J.A. Dehne, X. Wang, D.M. Wilder, and A. Barkatt. Effects of salts on silicateglass dissolution in water: kinetics and mechanisms of dissolution and surface cracking. Physics andChemistry of Glasses 40(3): 157-170. 1999.
12-43
13 PHASE STABILITY
The phase stability of glasses used forthe immobilization of HLW is of concern during melting and long-termstorage. The requirements for phase stability have been outlined in the WAPS. Phase instability in glass canbe induced by either liquid-liquid phase separation or crystallization on cooling from the melt. Whilecrystallization occurs only at temperatures belowthe T, liquid-liquid phase separation occurs in the fields ofimmiscibility defined by the equilibrium or metastable phase relations, which may be above or below theliquidus. These processes have been carefully controlled in many engineered glass products to achieveproperties not available in a homogeneous glass.
* High-silicaVycori glass is made at lower temperatures than an equivalent fused silica by a processinvolving phase separation ofa sodium borosilicate glass, chemical leaching ofthe high-alkali phase,and viscous sintering of the remaining high-silica network.
* Pyroceramlm glass ceramics are first melted and formed as glasses and subsequently crystallized tocontain more than 95 percent by volume of crystalline phases in the lithium aluminosilicate systemthat lend a low thermal expansion coefficient and high degree of thermal shock resistance to theproduct.
For the glass waste form, however, phase separation and crystallization can result in the development of aninhomogeneous microstructure that may affect the reliability of the production process and productperformance. A time-temperature-transformation (mT') diagram is a useful method for defining the heattreatment conditions for crystallization, and for predicting the types and quantities ofphases that can form inHLW glasses. From the nose of a TI' curve, one is able to calculate the rate of cooling needed at eachtemperature to avoid a specific volume fraction of crystalline phases. This information can provide guidanceabout conditions to be avoided during processing of the waste glasses such as melter idling, and shipping,handling, and storage of canistered waste forms. An overview of the phase separation and crystallizationbehavior in HLW glasses and their effect on glass performance is the subject of this chapter.
13.1 PHASE STABILITY SPECIFICATION
Specification 1.4 of the WAPS requires that MTT diagrams be developed for each of the targetedHLW glass compositions. The TTT diagrams identify the temperatures and durations of exposure at thesetemperatures that cause significant changes in phase structure and phase composition of the glass wasteform.
13.2 PHASE SEPARATION
Liquid-liquid phase separation is the growth of two or more noncrystalline glassy phases, each ofwhich has a different composition from the overall melt. The occurrence of phase separation is confinedwithin aregion of composition and temperature inthe phase diagram known as the immiscibility dome phaseboundary, as shown in figure 13-1. The dashed line delineates the metastable extension of the immiscibilityboundary. Single-phase glasses with ahomogeneous composition withinthe phase boundary can lower theirfree energyby separating into two immiscible glass phases. The compositions of these two glass phases canbe determined by drawing a common tangent between the energy minima of the free energy versus
13-1
a.~)
a) Li q |lQ.Ea) 8 iq~uid B0)
Liquid AImmiscibility A A+B
A B
Figure 13-1. Liquid immiscibility in a binary eutectic diagram of components A and B. XBis themole fraction of B.
composition curve. There are two types of phase separation kinetics: nucleation/growth and spineldecomposition. While the former is large in composition fluctuation but small in spatial extent, the latter is
small in composition fluctuation and large in spatial extent. These two processes often result in differentmicrostructures. Tomozawa (1979) classified the microstructure of the phase-separated glasses into threetypes, depending on the separation kinetics governed by nucleation/growth or spinel decomposition:
* Interconnected microstructure* Chemicallymore durable phase dispersed as discrete droplets in a continuous matrix ofthe less
durable phase* Chemically less durable phase dispersed as discrete droplets in a continuous matrix ofthe more
durable phase
Borosilicate glasses are prone to liquid-liquid phase separation. The phase-separated glass usuallyforms one phase rich in silica and a second one enriched in most ofthe other components. The latter is usuallyless durable than either the silica-rich phase or the homogeneous glass. Depending on the developedmicrostructure, phase separation can profoundly affect glass durability. Controlling the glass chemistry incompositional regions that avoid phase separation is the keyto achieving waste glass durability control duringprocessing.
13-2
13.2.1 Immiscibility Boundary Measurements
Immiscibility boundaries can be mapped on phase diagrams by the visible opalescence in mostphase-separated glasses. If a glass changes from transparent to uniformly opalescent after heating at a giventemperature for sufficient time, this indicates that the heat treatment temperature used is inside theimmiscibilityboundary. For a given glass composition, the maximum heat-treatment temperature at which theopalescence is detected can be defined as the critical (or immiscibility) temperature. A series of heattreatments on selected glass compositions can thus be used to map the immiscibility boundary. Care is neededto assure the opalescence is indeed caused by liquid-liquid phase separation rather than crystallization. Theaccuracy of the immiscibility boundary using the opalescence method is approximately ±10 'C. A moreaccurate determination can be made by checking the existence ofthe microstructure in the quenched sampleusing various instruments such as electron microscopy, X-ray diffraction, and the small angle neutronscattering method.
13.2.2 Phase Separation in High-Level Waste Glasses
Taylor (1990) suggested a simple approach to predicting immiscibility in complex borosilicatesystems with a special reference to nuclear waste immobilization. His approach is based on the observationthatthe extent of immiscibilityis related to the polarizing power ofthe modifier cation, which is a function ofits charge and radius. In general, as the polarizing power increases, the attraction between the cation innegatively charged nonbridging oxygens strengthens, and the tendency toward clustering ofthese species andphase separation increases. Figure 13-2 shows miscibilitylimits in avariety of binary silicate systems involvingmonovalent and divalent cations. As expected, the extent ofthe immiscibility dome increases with increasingpolarizing power. This relationship betweenthe extent of immiscibility and polarizing power breaks down withthe most polarizing cation (i.e., A13+, which enters the silicate structure as a network-former rather thana modifier). Consequently, a suppression of liquid-liquid phase separation results when sufficient A1203 isincorporated into the glass composition. To a lesser extent, additions ofTiO2 and ZrO2 will also suppress thedevelopment of liquid-liquid phase separation. Taylor used the location ofthe Na2O-B2 0 3-SiO2 submixturefrom the multicomponent glass from the immiscibility dome to predict the development of liquid-liquid phaseseparation. The effects of individual components on the occurrence of phase separation in a variety ofmulticomponent borosilicate glasses have been reviewed byTaylor (1990). Based on the observations of thisreview report, there are a large number of oxides of cations more polarizing than Na+, such as oxides of Li,Mg, Ca, Ba. Taylor predicted that the addition of various oxides will induce phase separation when added toa submixture composition lying within the sodium borosilicate immiscibility dome but will not promote phaseseparation when added to a submixture composition lying outside the dome.
Hrma et al. (1994) conducted a composition variation studyto characterize the relationships betweenglass composition and properties of a total of 123 glasses in support of the HLW glasses development at theHanford site. The authors applied Taylor's model to the Hanford simulated HLW glasses for the predictionof liquid-liquid phase separation in complex, multicomponent systems. Figure 13-3 shows the durability dataof all the tested glasses within various normalized sodium borosilicate submixtures with an immiscibilityboundary for the Na2O-B2 03-SiO2 system. In all three submixtures, glasses located near the SiO2-rich apexwere generally more durable because of a continuous 3D network, whereas glasses positioned in theSiO2-deficient regions produced low durability. The use of the normalized sodium borosilicate submixture(figure 1 3-3a) appears to be inadequate to predict phase separation for Hanford's glasses, in which many of
13-3
0 15000
a)I-
a)a-Eg 1 000
0 10 20 30 40
MO or X20 (mol%)
Figure 13-2. Miscibility limits in a variety of binary silicate systems involving monovalent (X+) anddivalent (MNV) cations (Kawamoto and Tomozawa, 1981)
the high-durability, nonphase-separated glasses remained within the immiscibilityboundary. With the additionof Li20 to the alkali corner of the sodium borosilicate submixture (figure 13-3b), the experimental resultsindicated that phase separation can be more accurately predicted for the multicomponent systems. Aspredicted by the modified model, nine glasses located within the immiscibility boundary were prone toliquid-liquid phase separation. The sodium borosilicate submixture with an equivalent Na20 content(figure 1 3-3c) further improved the abilityto predict immiscibilitywithinthese glass systems. The equivalentNa2O content is defined as
Na2 O(equivalent) = k x Li2 0 + Na 2O (13-1)
wherek= 61.98/29.88, the ratio ofthemolecularmasses ofNa2O and Li20, respectively. In this equivalentsubmixture, onlyeight glasses remained within the immiscibilityboundary, four which were experimentallycharacterized to be phase-separated. The modified submixture systems, both the addition of Li20 on a massfraction basis and the equivalent Na2O content on a molar basis, demonstrated the capacityto discriminatethe phase-separated glasses from the homogeneous glasses. A few glasses, however, did not follow the
13-4
,0.8
14 ( - -4 - X. . ,1.00.6 0.4 0.2 0
B2 03 SO 2
(a)
Na2O (Equivalent)Na2 O + La20
0.4 0.4
1.0 1.00.6 0.4 0.2 0 0.6 0.4 0.2 0
B203 SiO2 B203 SiO2
(b) (c)
Figure 13-3. Prediction of phase separation using various normalized submixtures of (a) Na2O-B 203-SiO 2, (b) (Na2O+Li2 O)-B203-SiO2 , and (c) Na2O(equivalent)-B 2 03 -SiO2. The dashed line isthe immiscibility boundary. Solid points and open circles represent high- and low-durability glasses,respectively (Hrma et al., 1994).
13-5
modified models for predicting immiscibility. Evaluation of other submixtures for immiscibility prediction wastherefore suggested.
Jantzen et al. (2000) investigated the compositional nature of phase separation in HLW glasses in anattemptto improve the predictability ofthe glass durabilitymodel by eliminating phase-separated glasses fromthe DWPF databases. An analysis ofthe compositional differences of 110 homogeneous and phase-separatedglasses indicated that the phase-separated glasses were significantly lower in A12 03 while being somewhathigher in B203. Good discrimination between the homogeneous and phase-separated glasses was achievedby plotting the sum of the lighter density alkali oxide components versus the denser sludge components, asshown in figure 13-4. Fromthese results, a composition-dependent phase separation discriminator has beendefined. This discriminator function is mathematically derived based on a discriminant analysis of 110 wasteglasses in 14 component composition spaces, and is expressed as
- 1.6035x -5.6478y + 210.9203 < 0 (13-2)
where x andy are defined as the sum of the wt% of less dense and dense oxide components, respectively,and are expressed as
22
21 -
20 -* *
19 U ~~~~~~*WCP BATCH I
18 9U)\ *W_ T WCPPUREX\
17 -wcPBAH2C ARMOY *WCP BATCH 4 * *WCP HL 'VW WP BLEND I
E 16 - P BATCH 3E 16 4HANFORD GLSS 0244 470
GI 15 -DMS 6839
O YEA0 14 -* *E
C Homogeneous Glasses \a) 13 + HUM-9
Homogeneous WCP SHUM-14
Glasses *. a U12 Phase Separated 4 \
Glasses (Hanford)11 Phase Separated *
10 Glasses (IDMS PX-4,-5,-6)
68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
Less Dense Oxide Components (wt%)
Figure 13-4. Compositional distinction between homogeneous and phase-separated glasses with95-percent confidence ellipsoids (Jantzen et al., 2000)
13-6
x = X wt%(Na 2 0 + Li 20 + K 2 0 + Cs02 + SiO2 + B 2 0 3 )
y =Ewt%(A 2°3 + Fe20 3 + Nd203 + Ce203 + La 2 03 + Y2 0 3 + CaO + MoO 3 )
Glasses willbehomogeneousifthe criterion defined inEq. (13-2) is satisfied. The phase separationdiscriminatorhas beenvalidated with a 95-percent accuracybyapplyingitto agroup of53 validation glassesthat cover a wide composition range.
13.3 CRYSTALLIZATION
Glass-crystal transformation follows atwo-step process: crystal nucleation and growth. When a liquidis cooled below its freezing point, crystallization occurs bythe growth of crystals at a finite rate from a finitenumber of nuclei. Glass formation may be attributed to a low rate of crystal growth, a low rate of nucleiformation, or a combination of both. The nucleation rate and crystal growth rate as a function oftemperaturecan be plotted in the form of skewed bell-shaped curves, as shown in figure 13-5. In both cases, the ratereaches a maximum at a certain temperature below the melting point and decreases to zero at both ends ofhigh and low temperatures. However, the maximum crystal growth rate occurs at much smaller undercoolingsthan the case for the maximum nucleation rate. The extent ofthe overlap between the two curves determinesthe capability of glass formation upon cooling: the smaller the overlap, the easier the glass formation.
TM
0.EU)
Metastabie Zoineof SupFeRooling __
-. 4--Crystal Growth Rate
Nucleation Rate
0 Rates of Nucleation and Growth
Figure 13-5. Variation of nucleation rate and crystal growth rate with temperature. T, representsmelting point.
13-7
A stable crystalline phase will onlyfonnwhenthe liquid is sufficientlyundercooled. The metastablezone of supercooling near the melting point with small undercoolings in figure 13-5 represents a region wherefew nuclei are present and the rate of crystal growth is low because of a reduced driving force forcrystallization. Contrarily, at large undercoolings belowthe melting point the nucleation rate maybe veryhighbut, in the absence of crystal growth, the overall crystallization rate remains low. A combination of nucleationrate and crystal growth rate is required for crystallization. Crystallization of nuclear waste glasses can occurduring continuous cooling of pour canisters, or during annealing of quenched glasses. The degree ofcrystallinity and crystalline phases that maybe produced will depend on factors such as the specific coolingrates and the waste glass compositions.
Glass crystallization affects both waste glass processability and acceptability. The formation ofcrystalline phases affects chemical durability, thus influencing the acceptability of the waste glass product forisolation in a geological repository. Crystallization can affect chemical durability as a result of a change ofresidual glass composition and structure. If the crystalline phases are more durable than the glass, thechemical durability of the partly crystallized glass is determined predominantly by the durability of residualglass. Crystalline phases in the melter can also cause processing problems, such as sludge formation on themelter bottom or melter electrode shorting if the phases are highly conductive. The TTT diagram is a usefulmethod for definingthe heattreatment conditions for crystallization, and for predicting the types and quantitiesof phases that can form in nuclear waste glasses. By combining the time and temperature dependencies ofthe nucleation and growth rates, the TFT diagram can be determined. The development of TTl diagrams forHLW glasses is a requirement of the WAPS.
13.3.1 Crystallization Kinetics Model
A quantitative model for the kinetics of crystallization can be developed by combining the phasetransformation theory of nucleation and growth based on the Johnson-Mehl-Avrami equation. Reynolds andHrma (1994) established an analytical kinetic model for spinel crystallization in a blended Hanford Site waste.The spinel volume fraction in molten glass, C, can be expressed in the form of the Johnson-Mehl-Avramiequation
-1=1 - exp[-(- ] (13-3)
where
CO - equilibrium valueI -time
- time constantn - Avrami exponent
Both CO and r are functions of temperature and expressed as
co = 1- exp[-BL( I - )] (13-4)Cmax T TL
13-8
T = Toexpj-yT)
where BL, B1, C,,., and r0 are constants, and TL is the liquidus temperature. The above kinetic andthermodynamic parameters can be determined by fitting the vol% of crystallization data, and then be used toconstruct the TTT diagram, as discussed in section 13.3.3.
13.3.2 Time-Temperature-Transformation Diagram Measurements
A MTT diagram can be determined experimentally by means of isothermal and nonisothermal heattreatments. For isothermal heat treatments, glass samples are heat-treated at specific temperatures for certaintime periods and then quenched to reserve the microstructure. Generally, the heat treatment temperaturesrange from Tto TL. Nonisothermal heattreatments involve a simulation ofcanister cooling curves and criticalcooling. The canister cooling curve simulates the thermal history of glasses in the canister, while criticalcooling simulates the glass at a specific cooling rate. In general, because the time required for crystallizationto begin and end is delayed during nonisothermal heat treatments, the isothermal diagrams tend to shift tolonger times and lower temperatures.
A variety of analytical techniques has been used to characterize the crystal phases in the heat-treatedglass samples, including optical microscopy, electron microscopy, and X-ray diffraction. The structure andcomposition of the crystalline phases can be identified by X-ray diffraction and an energy dispersivespectroscopy attachment to scanning/transmission electron microscopy. The vol% crystallization data of glasssamples are generallymeasured byimage analysis using reflected-light optical microscopy. With these vol%crystallization results, a TTT diagram can be constructed.
13.3.3 Time-Temperature-Transformation Diagrams for High-Level Waste Glasses
TTr diagrams for various waste glass compositions have been developed to determine the effect ofcomposition on crystallization. Bickford and Jantzen (1986) reported TIT diagrams for simulated glasses(SRS 131 and SRS 165) bounding the compositional range in the DWPF. Formulations included all of theminor constituents such as RuO2 and chromium, which have limited solubility in borosilicate glasses.Crystallization of these waste glasses under isothermal heat treatments was observed to be heterogeneousnucleation of spinel on RuO2 and subsequent nucleation of acmite on spinel. A similar crystallization path wasalso observed inthe glasses subjected to a canister centerline cooling heat treatment. Thetotalvol% of spinelin these glass samples, however, was slightly higher than those on isothermal annealing.
The lTlIl diagrams for seven HLW waste glass target compositions to be produced in the DWPFhave been determined by Cicero et al. (1993). The glass samples were isothermally heat-treated attemperatures varying from 500 to 1,100 0C withtimes varying from 0.75 to 768 hr. All glass compositionscontainedtrevorite, acmite, lithiummetasilicate, andnephelineinvariousamounts, exceptthePPUREXglasses,which did not contain any lithium metasilicate. Different phase regions in the TTI diagrams were thenconstructed according to the occurrence of each phase inthe time-temperature domain. Figure 13-6 showsa representative TTT diagram for the Blend composition of the DWPF glasses. As displayed in the figure,the trevorite region was predominant at higher temperatures. Acmite began to form with the trevorite from750 to 850 'C above 0.75 -hr annealing times and led to the ingrowth oflithium metasilicate at longer annealing
13-9
° 800- + AC A A Liq + TR + AC+ LMS
_ - \ a~~~+ *Licq + AC * Liq + AC. V- t *~~~~0+ + +LMS + +NE +LMS
D- 600 -
400 - TR = TrevoriteAC = Acmite
_ NE = NephelineLMS = Lithium Metasilicate
200.1 1 10 100 1000
Time (hr)
Figure 13-6. Time-temperature-transformation diagram for the Defense Waste Processing FacilityBlend glass (Cicero et al., 1993)
times. The normal growth sequence of acmite, lithiummetasilicate, and thennepheline occurred at the lowertemperatures.
Joseph et al. (1988) established TTT diagrams for the fully simulated reference borosilicate glass(WVCM50), both in the fully oxidized and partially reduced forms, developed for the WVDP. While theoxidized glass sample was directlymelted in air, the reduced glass was subsequentlyprocessed in a controlledreducing atmosphere. Crystallization behavior of these glasses was investigated using glass samplesisothermallyheat-treated overthetemperature range of 500 to 1,000 'C fortime periods of 1 to 384 hr. Themajor crystalline phases observed inthe heat-treated samples were spinel, acmite, and cerium-thorium oxide.The total vol% crystallization data, determined by image analysis, were used to construct TT diagrams atvarious levels of vol% crystals.
Simpson et al. (1994) constructed =TT diagrams forthe oxidized and reduced versions ofthe WVDPtarget Reference 6 glass with the consideration of statistical confidence intervals. A 95-percent confidenceinterval for the time to achieve a given vol% crystallization inthe region ofthe nose ofthe lTTl diagrams wasestimated. Spinel was identified in all of the glass samples. The reduced glass showed a greater extent ofcrystallizationthan the oxidized glass. The vol% crystallization in both glasses, however, was very low anddid not affect the processing of the WVDP Reference 6 glass. Figure 13-7 shows the l diagrams for theoxidized WVDP Reference 6 glass for 2 and 4 vol% crystallization. The 95-percent confidence intervals forthe time periods to achieve the respective crystallization are shown as horizontal bars.
13-10
500 1 1 1 1 1 I-2 -1 0 1 2 3 4
Log Time (hr)
Figure 13-7. Time-temperature-transformation diagrams for 2 and 4 vol% crystallization of theoxidized West Valley Demonstration Project Reference 6 Glass. Horizontal bars represent95-percent confidence intervals (Simpson et al., 1994).
Using the Johnson-Mehl-Avrami Eq. (13-3), Reynolds and Hrma (1997) obtained the kineticparameters by fitting isothermal data for spinel crystallization of a simulated HLW glass for the Hanford site.From these numerical values of coefficients, the T`TT diagram for isothermal crystallization was calculatedas shown in figure 13-8 a. In a similar way, nonisothermal kinetics of spinel crystallization during cooling atconstant rates was also predicted by Casler and Hrma (1999). The TTr diagram for nonisothermalcrystallization is shown in figure 1 3-8b. A comparison of the m diagrams was made by overlapping theisothermal TIT curve for 0.004 volume fraction ofspinel inthe nonisothermal diagram (figure 1 3-8b). Thenose ofthe nonisothermal curve was slightly shifted to a longer time and lower temperature than that for theisothermal curve.
13.4 EFFECT OF PHASE STABILITY ON GLASS DURABILITY
13.4.1 Effect of Phase Separation
Liquid-liquid phase separation has been shown to be detrimental to the stability and durability ofnuclear waste glasses (Hrma et al., 1994; Tovena et al., 1994). From the composition variation studyperformed byHrma et al. (1994), liquid-liquid phase separation was assessed using the plots ofthe MCC-1durabilitydatawithin various normalized submixtures, as shownin figure 13-1. Glasses withinthe immiscibilityboundary generally showed low durability. These low-durability glasses were characterized by eitherliquid-liquid phase separation or by a high degree of crystallinity.
13-11
10271100U = U.UU5 927 0.1
1~ 000=(C 0.01 ^ 827 C0.00 1
° 00 go - l 0 i.019CD =3~~~~~ 727-
= 800 \\ 0=0.02 , 627 0 0.022co\ a) 627 - l
0L 700 E 527
H 600 _0\\\C = 0.03 427 C =0.004
500 a | &\\\327 (T= const.)102 103 104 105 106 10710 13 14 15 6
Time (s) Time (s)
(a) (b)
Figure 13-8. Time-temperature-transformation diagrams for (a) isothermal and (b) nonisothermalcrystallization for a simulated Hanford glass with various volume fraction of spinel (Casler andHrma, 1999)
Tovena et al. (1994) reported dissolution rates of HLW glasses developed in France. Experimentalanalysis indicated that phase separation also resulted in a high glass dissolution rate. In addition, glassdurability ofthese glasses was assessed using various thermodynamic and structural models in the literature.The calculated dissolution rates were largely underestimated for the phase-separated glasses. Theunderprediction of the glass durability models was attributed to the development of inhomogeneousmicrostructure in the phase-separated glasses.
13.4.2 Effect of Crystallization
Glass durability can be increased, decreased, or unaffected by crystallization, depending on the typeand fraction of crystalline phases formed. Jantzen and Bickford (1985) investigated the effect ofcrystallization on the durability of Savannah River defense waste glasses. The crystalline phases precipitatedfrom the glasses after simulated canister centerline cooling were spinel, acmite, and nepheline. The resultsfromthe MCC-1 durabilitytests showed thatthenoncrystallized glass samples gave reproducible leach rates,but the crystallized samples did not. Variations of up to two times were observed for the leach rates ofcrystallized samples. Waste glasses that contained crystalline phases had leach rates one to three timesgreater than the noncrystallized glasses of the same composition.
Jain et al. (1993) reported the influence of heat treatment on the glass durability of the targetReference 6 composition produced at the WVDP. Spinel phases were identified as the major crystalline
13-12
phases in the glass after isothermal heat treatments at temperatures of 500 to 900 'C for periods of 1 to96 hr. The PCT results indicated that, in all cases, the boron leach rate of the heat-treated glass increased.For samples heat-treated at 600 and 700 0C, boron release increased with heat treatment time because ofthe increase of the volume fraction of spinel phases.
Glass durability as a function of crystallization was studied for Hanford HLW glasses subjected tothe canister centerline cooling heat treatment (Hrma et al., 1994). The crystalline phases occurring mostfrequently were spinel, Li2SiO3, and clinopyroxene, followed by olivine, SiC 2, hematite, zircon, orthopyroxene,and nepheline. Crystallization ofnepheline and SiO2 significantly decreased glass durability, and crystallizationof clinopyroxene and zircon showed almost no effect on glass durability.
13.5 TANK WASTE REMEDIATION SYSTEM CONCERNS
The phase stability of HLW glasses is sensitive to the overall glass composition and the thermalhistory ofthe system. Both liquid-liquid phase separation and crystallization have detrimental effects on theglass durability. Liquid-liquid phase separation can be suppressed by controlling the polarizing power ofvarious additive oxides, and the heat treatment conditions for crystallization can be specifically determinedthrough the development of a ITI' diagram. A design of the target glass composition and a heat-treatmentprocess of the waste glass that prevents significant phase changes to ensure glass durability is desirable tomeet the requirements of the WAPS.
13.6 SUMMARY
Phase stability in glass can be affected by either liquid-liquid phase separation or crystallization oncooling from melt. Forthe glass waste form, boththe phase separation and crystallization processes can resultininhomogeneous microstructure in glass that may affect the reliability ofthe waste glass process and productperformance. The phase stability requirements that include the development of a TIT diagram for eachprojected waste type have been outlined inthe WAPS. Liquid-liquid phase separation can be confined withinan immiscibilityboundary, which can be experimentally determined using the opalescence method. Varioussodium borosilicate submixture systems have been successfullyused to predict phase separation in HanfordHLW glasses. In an attempt to improve the predictability of the glass durability model, acomposition-dependent phase separation discriminatorhas been mathematicallyderived and used to distinguishthe phase-separated glasses from the homogeneous glasses. Crystallization can occur in nuclear wasteglasses during continuous cooling of pour canisters and annealing of quenched glasses. The TTT diagram isauseful tool for scoping theheattreatment conditions for crystallization. TTT diagrams for each ofthe majorHLW glasses have been developed. Experimental results indicate that both liquid-liquid phase separation andcrystallization have a detrimental effect on the glass durability ofnuclear waste glasses. Controlling the glasscompositions and defining the heat treatment conditions that avoid phase separation and crystallization arekeys to achieving waste glass durability control during processing.
13-13
13.7 REFERENCES
Bickford, D.F., and C.M. Jantzen. Devitrification of defense nuclear waste glasses: Role of melt insolubles.Journal of Non-Crystalline Solids 84: 299-307. 1986.
Casler, D.G., and P.R. Hrma. Nonisothermal kinetics of spinel crystallization in HLW glass.Proceedings of the Materials Research Society Symposium. Symposium Proceedings 556.Pittsburgh, PA: Materials Research Society: 255-262. 1999.
Cicero, C.A., S.L. Marra, and M.K. Andrews. Phase Stability Determinations of DWPF WasteGlasses (U). WSRC-TR-93-227. Revision 0. Aiken, SC: Westinghouse Savannah River Company.1993.
Hrma, P.R., G.F. Piepel, M.J. Schweiger, D.E. Smith, D.-S. Kim, P.E. Redgate, J.D. Vienna, C.A. LoPresti,D.B. Simpson, D.K. Peeler, and M.H. Langowski. Property/Composition Relationships forHanford High-Level Waste Glasses Melting at 1150 'C. Volume 2. PNL-10359. CeramicTransactions Richland, WA: Pacific Northwest National Laboratory. 1994.
Jain, V., S.M. Barnes, I. Joseph, L.D. Pye, I.L. Pegg, and P.B. Macedo. Impact of heat-treatment on thechemical stability of nuclear waste glasses produced at the West Valley Demonstration Project(WVDP). Proceedings of the Glass and Optical Materials Division Meeting of the AmericanCeramic Society, Stone Mountain, Georgia. Ceramic Transactions Volume 30.M.C. Weinberg, ed. Westerville, OH: American Ceramic Society: 367-370. 1993.
Jantzen, C.M., K.G. Brown, and T.B. Edwards. Predicting phase separation in nuclear waste glasses.Proceedings of the Environmental Issues and Waste Management Technologies in the Ceramicand Nuclear Industries V, Indianapolis, Indiana, April 25-28, 1999. Ceramic TransactionsVolume 107. G. Chandler and X. Feng, eds. Westerville, OH: American Ceramic Society. 2000.
Jantzen, C.M., and D.S. Bickford. Leaching of devitrified glass containing simulated SRP nuclear waste.Proceedings of the Materials Research Society Symposium. Symposium Proceedings 44.Pittsburgh, PA: Materials Research Society: 135-146. 1985.
Joseph, I., A. Mathur, C. Capozzi, J. Sehgal, D. Butts, D. McPherson, and L.D. Pye. The CrystallizationBehavior of the West Valley Reference Borosilicate Glass. DOE-NE-44139-44.West Valley,NY: West Valley Nuclear Services Company, Inc. 1988.
Kawamoto, Y., and M. Tomozawa. Prediction of immiscibilityboundaries ofternary silicate glasses. PhysicsChemistry Glasses 22: 11-16. 1981.
Reynolds, J.G., and P.R. Hrma. The kinetics of spinel crystallization from a high-level waste glass.Proceedings of the Materials Research Society Conference. Symposium Proceedings 465.Pittsburgh, PA: Materials Research Society: 65-69. 1997.
13-14
Simpson, J.C., D. Oksoy, T.C. Cleveland, L.D. Pye, and V. Jain. The statistics of the time-temperature-transformation diagram for oxidized and reduced West Valley reference 6 glass. Proceedings ofthe Environmental and Waste Management Issues in the Ceramic Industry II Symposium.96th Annual Meeting of the American Ceramic Society, Indianapolis, Indiana,April 25-27, 1994. Ceramic Transactions Volume 45. D. Bickford, S. Bates, V. Jain, andG. Smith, eds. Westerville, OH: American Ceramic Society: 377-387. 1994.
Taylor, P. A Review of Phase Separation in Borosilicate Glasses, With Reference to Nuclear WasteImmobilization. AECL-10173. Pinawa Manitoba, Canada: Whiteshell Nuclear ResearchEstablishment. 1990.
Tomozawa, M. Phase Separation in glass. In Treatise on Materials Science and Technology. CeramicTransactions Volume 17. M. Tomozawaand R.H. Doremus eds., NewYork: Academic Press, Inc.:71-113. 1979.
Tovena, I., T. Advocat, D. Ghaleb, E. Vemaz, and F. Larche. Thermodynamic and structural modelscompared with the initial dissolution rates on "SON" glass melts. Proceedings of the MaterialsResearch Society Conference. Symposium Proceedings 333. Pittsburgh, PA: Materials ResearchSociety: 595-602. 1994.
13-15
14 SUMMARY
The review of the history of waste form development and the information (data and models) on glass meltproperties for various type of commercial and waste glass compositions indicate that the behavior of glassesis complex and cannot be expressed in simple terms for all wastes. The glass composition needs to bedesigned uniquely for each waste type. The lessons learned or the existing database of HLW glasses can beused as a guideline to narrow the scope of development research needed to meet the requirements.
This report has been written to assist the NRC in
* Determining if sufficient information exists to assess safety considerations regarding vitrification of theHanford wastes
* Determining if current regulatoryguidelines are adequate for controlling implementation of vitrificationprocess to immobilize HLW at Hanford
* Identifying the existence of technical uncertainties
* Assessing where future guidance may be warranted.
14-1